diff --git a/.github/workflows/ci-report.yml b/.github/workflows/ci-report.yml
index 03908bb2b..03c933ad0 100644
--- a/.github/workflows/ci-report.yml
+++ b/.github/workflows/ci-report.yml
@@ -95,31 +95,33 @@ jobs:
 
           # Validate PR number is a positive integer (artifact comes from
           # untrusted fork code, so treat contents defensively).
-          PR_NUM=$(cat "$DIR/pr_number" | tr -d '[:space:]')
+          PR_NUM=$(tr -d '[:space:]' < "$DIR/pr_number")
           if ! [[ "$PR_NUM" =~ ^[0-9]+$ ]]; then
             echo "skip=true" >> "$GITHUB_OUTPUT"
             echo "::error::Invalid PR number in artifact: '$PR_NUM'"
             exit 0
           fi
 
-          echo "skip=false"              >> "$GITHUB_OUTPUT"
-          echo "pr_number=$PR_NUM"       >> "$GITHUB_OUTPUT"
           # Validate job-result strings against known GitHub Actions values.
           # Artifact contents come from the PR workflow (potentially untrusted
           # fork code), so we whitelist to prevent newline injection into
           # GITHUB_OUTPUT.
           validate_result() {
             local val
-            val=$(cat "$1" | tr -d '[:space:]')
+            val=$(tr -d '[:space:]' < "$1")
             case "$val" in
               success|failure|cancelled|skipped) echo "$val" ;;
               *) echo "unknown" ;;
             esac
           }
 
-          echo "quality=$(validate_result "$DIR/quality_result")"        >> "$GITHUB_OUTPUT"
-          echo "tests=$(validate_result "$DIR/tests_result")"            >> "$GITHUB_OUTPUT"
-          echo "e2e=$(validate_result "$DIR/e2e_result")"                >> "$GITHUB_OUTPUT"
+          {
+            echo "skip=false"
+            echo "pr_number=$PR_NUM"
+            echo "quality=$(validate_result "$DIR/quality_result")"
+            echo "tests=$(validate_result "$DIR/tests_result")"
+            echo "e2e=$(validate_result "$DIR/e2e_result")"
+          } >> "$GITHUB_OUTPUT"
 
       - name: Checkout (for vitest config)
         if: steps.meta.outputs.skip != 'true'
@@ -279,14 +281,17 @@ jobs:
             fi
           }
 
-          read CLI_T CLI_P CLI_F CLI_S CLI_SU CLI_D <<< "$(sum_results "$RESULTS_FILE")"
-          read WEB_T WEB_P WEB_F WEB_S WEB_SU WEB_D <<< "$(sum_results "$WEB_RESULTS_FILE")"
+          # `_` placeholder for the suite-count column — positional
+          # readability for sum_results' 6-field output, but the value
+          # isn't surfaced in the report (suites are tracked per-test
+          # framework, not as a top-line metric).
+          read -r CLI_T CLI_P CLI_F CLI_S _ CLI_D <<< "$(sum_results "$RESULTS_FILE")"
+          read -r WEB_T WEB_P WEB_F WEB_S _ WEB_D <<< "$(sum_results "$WEB_RESULTS_FILE")"
 
           TOTAL=$((CLI_T + WEB_T))
           PASSED=$((CLI_P + WEB_P))
           FAILED=$((CLI_F + WEB_F))
           SKIPPED=$((CLI_S + WEB_S))
-          SUITES=$((CLI_SU + WEB_SU))
           DURATION=$((CLI_D > WEB_D ? CLI_D : WEB_D))
 
           # ── Status helpers ──
diff --git a/.github/workflows/pr-autofix-publish.yml b/.github/workflows/pr-autofix-publish.yml
new file mode 100644
index 000000000..a22f2cc7a
--- /dev/null
+++ b/.github/workflows/pr-autofix-publish.yml
@@ -0,0 +1,314 @@
+name: PR Autofix (publish)
+
+# TRUSTED HALF of the autofix pipeline.
+#
+# Triggered by `pr-autofix.yml` completing on a PR (including fork PRs).
+# Downloads the diff artifact produced by the untrusted job and posts
+# inline review-comment suggestions to the PR using `reviewdog`. This
+# job NEVER checks out fork code — it only consumes the diff (data) and
+# calls the GitHub API. That isolation is what makes it safe to run
+# under `pull-requests: write` on fork-triggered events.
+#
+# Also posts (or edits) a single sticky summary comment so contributors
+# and AI agents have one stable, machine-readable signal that says
+# whether autofix had anything to suggest. Look for the heading
+# "## :sparkles: PR Autofix" in the PR's top-level comments.
+#
+# Reviewdog reporter: `github-pr-review` reads $REVIEWDOG_GITHUB_API_TOKEN
+# and posts via the GraphQL/REST PR-review API. It does not need a
+# checkout because the diff itself encodes file paths + line numbers.
+
+on:
+  workflow_run:
+    workflows: ['PR Autofix']
+    types: [completed]
+
+concurrency:
+  # Key on PR identity, NOT workflow_run.id — workflow_run.id is per-run
+  # unique, which would defeat serialization and let two parallel
+  # publishes both POST a sticky summary comment. CONTRIBUTING.md
+  # § GitHub Actions — Concurrency Convention names this anti-pattern
+  # explicitly. For fork PRs, `pull_requests[]` is empty in the
+  # workflow_run payload, so we fall back to head-repo + head-branch.
+  group: ${{ github.workflow }}-${{ github.event.workflow_run.pull_requests[0].number || format('{0}/{1}', github.event.workflow_run.head_repository.full_name, github.event.workflow_run.head_branch) }}
+  cancel-in-progress: false
+
+permissions: {}
+
+jobs:
+  publish:
+    name: publish-autofix
+    if: >-
+      github.event.workflow_run.event == 'pull_request'
+      && github.event.workflow_run.conclusion == 'success'
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    permissions:
+      pull-requests: write
+      # Required by actions/download-artifact to fetch artifacts produced
+      # by a different workflow run.
+      actions: read
+      # Required to create the `gitnexus/autofix` Check Run that reports
+      # the outcome (clean / suggestions-posted / skipped-too-large) to
+      # the PR's Checks tab. Branch protection or agents can grep the
+      # conclusion + output title without parsing the sticky comment.
+      checks: write
+    steps:
+      # Pinned to v8.0.1. Verify SHA via:
+      #   gh api repos/actions/download-artifact/git/refs/tags/v8.0.1
+      - name: Download autofix artifact
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
+        with:
+          name: autofix
+          run-id: ${{ github.event.workflow_run.id }}
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          path: autofix-in
+
+      - name: Read and validate metadata
+        id: meta
+        shell: bash
+        run: |
+          set -euo pipefail
+          test -f autofix-in/metadata.json
+          jq . autofix-in/metadata.json
+
+          # The artifact comes from the untrusted half running fork code.
+          # Every field is allowlist-validated before it can flow into
+          # $GITHUB_OUTPUT. A newline in head_ref would otherwise let a
+          # malicious branch name inject a second `pr_number=N` line and
+          # redirect this job's reviewdog suggestions / sticky summary
+          # comment onto a victim PR under github-actions[bot] with
+          # pull-requests: write.
+          assert_field() {
+            local key="$1" pattern="$2" value
+            value=$(jq -r ".${key} // empty" autofix-in/metadata.json)
+            if [ -z "$value" ] || ! [[ "$value" =~ $pattern ]]; then
+              echo "::error::metadata.${key} failed allowlist (got: $(printf '%q' "$value"))"
+              exit 1
+            fi
+            printf '%s' "$value"
+          }
+
+          SCHEMA=$(assert_field    schema         '^gitnexus\.pr-autofix/v[0-9]+$')
+          PR_NUMBER=$(assert_field pr_number      '^[0-9]+$')
+          HEAD_SHA=$(assert_field  head_sha       '^[0-9a-f]{40}$')
+          HEAD_REF=$(assert_field  head_ref       '^[A-Za-z0-9._/-]+$')
+          HEAD_REPO=$(assert_field head_repo      '^[A-Za-z0-9._-]+/[A-Za-z0-9._-]+$')
+          BASE_REPO=$(assert_field base_repo      '^[A-Za-z0-9._-]+/[A-Za-z0-9._-]+$')
+          CHANGED=$(assert_field   changed_lines  '^[0-9]+$')
+
+          # Defence-in-depth: refuse to act if the artifact claims to
+          # belong to a different repo than the one that triggered us.
+          if [ "$BASE_REPO" != "${GITHUB_REPOSITORY}" ]; then
+            echo "::error::Artifact base_repo does not match \$GITHUB_REPOSITORY — refusing to publish."
+            exit 1
+          fi
+
+          {
+            echo "schema=${SCHEMA}"
+            echo "pr_number=${PR_NUMBER}"
+            echo "head_sha=${HEAD_SHA}"
+            echo "head_ref=${HEAD_REF}"
+            echo "head_repo=${HEAD_REPO}"
+            echo "base_repo=${BASE_REPO}"
+            echo "changed_lines=${CHANGED}"
+          } >> "$GITHUB_OUTPUT"
+
+      # Pinned to v1.5.0. Verify SHA via:
+      #   gh api repos/reviewdog/action-setup/git/refs/tags/v1.5.0
+      #   (annotated tag — resolve via .../git/tags/<sha> --jq .object)
+      - name: Install reviewdog
+        if: steps.meta.outputs.changed_lines != '0'
+        uses: reviewdog/action-setup@d8a7baabd7f3e8544ee4dbde3ee41d0011c3a93f # v1.5.0
+        with:
+          # Pin the binary, not just the action SHA — a bad reviewdog
+          # release otherwise breaks every PR with no rollback. Bump
+          # this knob deliberately when validating a new release.
+          reviewdog_version: v0.21.0
+
+      - name: Post inline suggestions
+        id: suggest
+        if: steps.meta.outputs.changed_lines != '0'
+        env:
+          REVIEWDOG_GITHUB_API_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          CI_REPO_OWNER: ${{ github.repository_owner }}
+          CI_REPO_NAME: ${{ github.event.repository.name }}
+          CI_PULL_REQUEST: ${{ steps.meta.outputs.pr_number }}
+          CI_COMMIT: ${{ steps.meta.outputs.head_sha }}
+          # Pull `changed_lines` through env so bash gets a real
+          # variable (and shellcheck SC2170 doesn't fire on `-gt` against
+          # a `${{ }}`-interpolated literal).
+          CHANGED_LINES: ${{ steps.meta.outputs.changed_lines }}
+        shell: bash
+        run: |
+          set -euo pipefail
+          patch=autofix-in/autofix.patch
+          if [ ! -s "$patch" ]; then
+            echo "Empty patch — nothing to suggest."
+            echo "posted=false" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+
+          # GitHub's review-comment API returns 406 on diffs above ~3k
+          # changed lines. Bail out gracefully and let the summary
+          # comment carry the signal instead.
+          if [ "$CHANGED_LINES" -gt 3000 ]; then
+            echo "Diff too large ($CHANGED_LINES lines) — skipping inline suggestions."
+            echo "posted=skipped-too-large" >> "$GITHUB_OUTPUT"
+            exit 0
+          fi
+
+          # `-f.diff.strip=1` matches `git diff` output (a/foo b/foo).
+          # `-filter-mode=added` only suggests on lines the PR added,
+          #   which avoids re-suggesting on already-resolved threads when
+          #   the contributor re-adds the autoformat label.
+          reviewdog \
+            -f=diff -f.diff.strip=1 \
+            -name="prettier+eslint" \
+            -reporter=github-pr-review \
+            -filter-mode=added \
+            -level=warning \
+            -fail-on-error=false < "$patch"
+
+          echo "posted=true" >> "$GITHUB_OUTPUT"
+
+      - name: Upsert sticky summary comment
+        # Only post when ci-quality found something fixable (= the
+        # autofix patch is non-empty). When prettier/eslint are clean
+        # the patch is zero bytes and the sticky comment is pure noise,
+        # so we skip it. When the diff was too large for inline
+        # suggestions, the sticky is the only signal the contributor
+        # gets, so we still post in that case.
+        if: >-
+          always()
+          && steps.meta.outputs.pr_number != ''
+          && steps.meta.outputs.changed_lines != '0'
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          GH_REPO: ${{ github.repository }}
+          PR: ${{ steps.meta.outputs.pr_number }}
+          CHANGED: ${{ steps.meta.outputs.changed_lines }}
+          HEAD_SHA: ${{ steps.meta.outputs.head_sha }}
+          SCHEMA: ${{ steps.meta.outputs.schema }}
+          POSTED: ${{ steps.suggest.outputs.posted }}
+          RUN_ID: ${{ github.run_id }}
+        shell: bash
+        run: |
+          set -euo pipefail
+
+          # Stable heading + marker — agents grep for these exact strings.
+          marker="<!-- gitnexus:pr-autofix-summary -->"
+          heading="## :sparkles: PR Autofix"
+
+          if [ "${POSTED}" = "skipped-too-large" ]; then
+            ui_state="skipped-too-large"
+            prose="Diff is **${CHANGED}** lines — too large for inline suggestions (GitHub caps the review-comment API at ~3000). Run locally: \`npm run lint:fix && npm run format\`."
+          else
+            ui_state="suggestions-posted"
+            prose="Posted formatting / unused-import suggestions inline. Click **Apply suggestion** on each, or run locally: \`npm run lint:fix && npm run format\`."
+          fi
+
+          # Machine-readable JSON block — agents parse this instead of
+          # regexing English. Fenced code-block info string is
+          # `gitnexus-autofix` so agents can locate it without ambiguity.
+          json=$(jq -n -c \
+            --arg schema       "${SCHEMA}" \
+            --arg state        "${ui_state}" \
+            --argjson pr_number "${PR}" \
+            --argjson changed_lines "${CHANGED}" \
+            --arg head_sha     "${HEAD_SHA}" \
+            --arg run_id       "${RUN_ID}" \
+            '{schema:$schema, state:$state, pr_number:$pr_number, changed_lines:$changed_lines, head_sha:$head_sha, run_id:$run_id}')
+
+          # Multi-line quoted string instead of a column-0 heredoc — YAML's
+          # `run: |` block ends as soon as a content line dedents below the
+          # block's first-line indent, which would mis-parse the workflow.
+          body="${marker}
+          ${heading}
+
+          ${prose}
+
+          \`\`\`gitnexus-autofix
+          ${json}
+          \`\`\`"
+          # Strip the leading 10-space indent that the YAML block requires
+          # so the rendered comment body starts at column 0.
+          body="$(printf '%s\n' "$body" | sed 's/^          //')"
+
+          # Small retry wrapper for transient 5xx / rate-limit responses
+          # on the GitHub REST API. Three tries with linear backoff. We
+          # only retry GET (idempotent) and PATCH on a known comment id
+          # (idempotent). POST is NOT wrapped — retrying a comment-create
+          # would create duplicates if the first attempt actually landed.
+          gh_retry() {
+            local n=0 max=3
+            while true; do
+              if gh "$@"; then return 0; fi
+              n=$((n+1))
+              if [ "$n" -ge "$max" ]; then return 1; fi
+              sleep $((n * 2))
+            done
+          }
+
+          # Find existing bot comment by the marker and edit-in-place; else create.
+          # CRITICAL: filter by `.user.login == "github-actions[bot]"`. A regular
+          # user posting a comment containing the marker would otherwise be the
+          # `head -n1` match; PATCH on someone else's comment 403s, `set -e`
+          # aborts, and the bot is permanently DoS'd for that PR.
+          existing=$(gh_retry api "repos/${GH_REPO}/issues/${PR}/comments" \
+            --paginate --jq ".[] | select(.user.login == \"github-actions[bot]\" and (.body | contains(\"${marker}\"))) | .id" \
+            | head -n1 || true)
+
+          if [ -n "${existing}" ]; then
+            gh_retry api -X PATCH "repos/${GH_REPO}/issues/comments/${existing}" \
+              -f body="${body}" >/dev/null
+            echo "Updated comment ${existing}."
+          else
+            gh api -X POST "repos/${GH_REPO}/issues/${PR}/comments" \
+              -f body="${body}" >/dev/null
+            echo "Created summary comment."
+          fi
+
+      - name: Emit gitnexus/autofix Check Run
+        # Stable check name `gitnexus/autofix` so PR-watching agents can
+        # `gh pr checks <pr>` and read the conclusion + title without
+        # parsing the sticky comment. Three outcomes:
+        #   clean              → conclusion: success
+        #   suggestions-posted → conclusion: neutral (review suggestions)
+        #   skipped-too-large  → conclusion: neutral (diff > 3000 lines)
+        # `neutral` does not block branch-protection required-checks but
+        # is visually distinct from a green pass.
+        if: always() && steps.meta.outputs.head_sha != ''
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          GH_REPO: ${{ github.repository }}
+          HEAD_SHA: ${{ steps.meta.outputs.head_sha }}
+          CHANGED: ${{ steps.meta.outputs.changed_lines }}
+          POSTED: ${{ steps.suggest.outputs.posted }}
+        shell: bash
+        run: |
+          set -euo pipefail
+
+          if [ "${CHANGED}" = "0" ]; then
+            conclusion="success"
+            title="Formatting clean"
+            summary="Prettier and ESLint --fix produced no changes."
+          elif [ "${POSTED}" = "skipped-too-large" ]; then
+            conclusion="neutral"
+            title="Diff too large for inline suggestions (${CHANGED} lines)"
+            summary="GitHub caps the review-comment API at ~3000 lines. Run \`npm run lint:fix && npm run format\` locally."
+          else
+            conclusion="neutral"
+            title="Suggestions posted"
+            summary="Inline review-comment suggestions posted. Click **Apply suggestion** on each, or run \`npm run lint:fix && npm run format\` locally."
+          fi
+
+          gh api -X POST "repos/${GH_REPO}/check-runs" \
+            -f name="gitnexus/autofix" \
+            -f head_sha="${HEAD_SHA}" \
+            -f status="completed" \
+            -f conclusion="${conclusion}" \
+            -f "output[title]=${title}" \
+            -f "output[summary]=${summary}" \
+            >/dev/null
+          echo "Posted check-run gitnexus/autofix=${conclusion} (${title})"
diff --git a/.github/workflows/pr-autofix.yml b/.github/workflows/pr-autofix.yml
new file mode 100644
index 000000000..f15e8c0e6
--- /dev/null
+++ b/.github/workflows/pr-autofix.yml
@@ -0,0 +1,146 @@
+name: PR Autofix
+
+# UNTRUSTED HALF of the autofix pipeline.
+#
+# Runs `npm run lint:fix` + `npm run format` against the PR head
+# (including fork heads) and uploads the resulting diff as an artifact.
+# This job has NO privileged token and CANNOT post to the PR. The trusted
+# `pr-autofix-publish.yml` workflow downloads the artifact via
+# `workflow_run` and posts the inline review-comment suggestions.
+#
+# Why the split:
+#   ESLint loads plugins from fork-controlled `node_modules`, so running
+#   it in a job with `pull-requests: write` would let a malicious fork PR
+#   ship a poisoned eslint plugin and execute arbitrary code under that
+#   token. By keeping fork code execution in this job (token: read-only)
+#   and posting from a separate trusted job that never touches fork
+#   code, we get the inline-suggestion UX for fork PRs without the
+#   supply-chain hole. (See autofix.ci for the same pattern.)
+#
+# Removes unused imports via `eslint-plugin-unused-imports`, already in
+# devDependencies and wired into the `lint` config.
+
+on:
+  pull_request:
+    types: [opened, synchronize, reopened]
+    # Skip lockfile / generated-file PRs entirely — `action-suggester`
+    # cannot post on diffs > ~3k lines (GitHub returns 406) and these
+    # paths produce massive diffs no human wants suggested back inline.
+    paths-ignore:
+      - '**/package-lock.json'
+      - '**/*.snap'
+      - '**/dist/**'
+      - '**/node_modules/**'
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number }}
+  # Don't cancel in-flight runs; the publish workflow may already be
+  # downloading the artifact and a cancelled untrusted run produces no
+  # signal at all (worse DX than waiting).
+  cancel-in-progress: false
+
+# This workflow runs untrusted fork code. Top-level deny-all and NO
+# job-level grants — the job can only read its own checkout.
+permissions: {}
+
+jobs:
+  autofix:
+    name: autofix
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    steps:
+      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          # PR head commit (not the synthetic merge ref) — we need the
+          # exact tree the contributor pushed so suggestions line up.
+          ref: ${{ github.event.pull_request.head.sha }}
+          repository: ${{ github.event.pull_request.head.repo.full_name }}
+          persist-credentials: false
+
+      - uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e # v6.4.0
+        with:
+          node-version: 20
+          cache: npm
+          cache-dependency-path: package-lock.json
+
+      # `--ignore-scripts` blocks pre/postinstall lifecycle hooks. ESLint
+      # plugins still load from node_modules (that is the actual escape
+      # hatch on a typical fork), but this job has no token to abuse —
+      # which is the whole point of the split.
+      - run: npm ci --ignore-scripts
+
+      - name: ESLint --fix (removes unused imports)
+        run: npm run lint:fix
+        # Lint errors that --fix can't auto-resolve must not block the
+        # diff artifact — partial fixes are still useful as suggestions.
+        continue-on-error: true
+
+      - name: Prettier --write
+        run: npm run format
+        continue-on-error: true
+
+      - name: Capture diff and metadata
+        id: capture
+        # Pass GitHub-context values via env: rather than `${{ }}`
+        # interpolated directly into the bash body. `head.ref` and
+        # `head.repo.full_name` are fork-controlled strings; expanding
+        # them into shell source is the canonical template-injection
+        # vector zizmor flags. Even though this job has `permissions: {}`,
+        # routing through env: makes it impossible for a future scope
+        # grant to turn into RCE. Inside bash, reference as `$HEAD_REF`
+        # etc. — the values are then plain strings, not code.
+        env:
+          PR_NUMBER: ${{ github.event.pull_request.number }}
+          HEAD_SHA: ${{ github.event.pull_request.head.sha }}
+          HEAD_REF: ${{ github.event.pull_request.head.ref }}
+          HEAD_REPO: ${{ github.event.pull_request.head.repo.full_name }}
+          BASE_REPO: ${{ github.repository }}
+        shell: bash
+        run: |
+          set -euo pipefail
+          mkdir -p autofix-out
+
+          # Produce a unified diff of the working tree vs. the PR head.
+          # Empty diff => nothing to suggest; the publish job short-circuits.
+          git diff --no-color > autofix-out/autofix.patch
+
+          # NOTE: `changed_lines` is the line-count of the patch file,
+          # which includes hunk headers and context lines — NOT the
+          # added/removed source-line count. The 3000-line cap in
+          # pr-autofix-publish.yml is therefore conservative (fires
+          # before reviewdog hits GitHub's ~3k review-comment API
+          # ceiling). That bias is intentional.
+          changed_lines=$(wc -l < autofix-out/autofix.patch | tr -d ' ')
+          echo "changed_lines=${changed_lines}" >> "$GITHUB_OUTPUT"
+
+          # Carry PR identity over to the trusted job. workflow_run
+          # context is base-repo-only, so the publish job needs these
+          # to call the GitHub PR API on the right resource.
+          # CONTRACT: keep this schema in sync with pr-autofix-publish.yml's
+          # `assert_field` validators and the agent-facing JSON block in
+          # the sticky comment. Bump `schema` when changing field names.
+          jq -n \
+            --arg     schema        'gitnexus.pr-autofix/v1' \
+            --argjson pr_number     "${PR_NUMBER}" \
+            --arg     head_sha      "${HEAD_SHA}" \
+            --arg     head_ref      "${HEAD_REF}" \
+            --arg     head_repo     "${HEAD_REPO}" \
+            --arg     base_repo     "${BASE_REPO}" \
+            --argjson changed_lines "${changed_lines}" \
+            '{schema:$schema, pr_number:$pr_number, head_sha:$head_sha, head_ref:$head_ref, head_repo:$head_repo, base_repo:$base_repo, changed_lines:$changed_lines}' \
+            > autofix-out/metadata.json
+
+          echo "--- metadata ---"
+          cat autofix-out/metadata.json
+          echo "--- diff (head) ---"
+          head -c 2000 autofix-out/autofix.patch || true
+
+      # Pinned to v7.0.1. Verify SHA via:
+      #   gh api repos/actions/upload-artifact/git/refs/tags/v7.0.1
+      - name: Upload autofix artifact
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
+        with:
+          name: autofix
+          path: autofix-out/
+          retention-days: 1
+          if-no-files-found: error
diff --git a/.github/workflows/release-candidate.yml b/.github/workflows/release-candidate.yml
index 9ab00c0bd..36f37b5c6 100644
--- a/.github/workflows/release-candidate.yml
+++ b/.github/workflows/release-candidate.yml
@@ -294,9 +294,11 @@ jobs:
             fi
           fi
 
-          echo "base=$BASE"             >> "$GITHUB_OUTPUT"
-          echo "rc_n=$NEXT_N"           >> "$GITHUB_OUTPUT"
-          echo "rc_version=$RC_VERSION" >> "$GITHUB_OUTPUT"
+          {
+            echo "base=$BASE"
+            echo "rc_n=$NEXT_N"
+            echo "rc_version=$RC_VERSION"
+          } >> "$GITHUB_OUTPUT"
 
       - name: Apply rc version in-CI
         shell: bash
@@ -354,9 +356,11 @@ jobs:
           # remote ref, the push fails and we stop before npm publish.
           git push --atomic origin "refs/tags/$VTAG" "refs/tags/$MARKER"
 
-          echo "vtag=$VTAG"     >> "$GITHUB_OUTPUT"
-          echo "marker=$MARKER" >> "$GITHUB_OUTPUT"
-          echo "release_sha=$RELEASE_SHA" >> "$GITHUB_OUTPUT"
+          {
+            echo "vtag=$VTAG"
+            echo "marker=$MARKER"
+            echo "release_sha=$RELEASE_SHA"
+          } >> "$GITHUB_OUTPUT"
 
       - name: Publish to npm (rc dist-tag)
         run: npm publish --provenance --access public --tag rc
diff --git a/.github/workflows/workflow-lint.yml b/.github/workflows/workflow-lint.yml
index 97f388d28..7b38b8ddd 100644
--- a/.github/workflows/workflow-lint.yml
+++ b/.github/workflows/workflow-lint.yml
@@ -1,10 +1,13 @@
-name: Workflow Lint (zizmor)
+name: Workflow Lint
 
-# Lints .github/workflows/** for known GitHub Actions security misconfigurations:
-# unpinned Actions, dangerous ${{ ... }} interpolation in run: blocks,
-# missing per-job permissions:, etc.
+# Lints .github/workflows/** for both:
+#   - actionlint: YAML syntax, expression typing, shellcheck inside `run:`
+#     blocks, unknown contexts, deprecated runner labels.
+#   - zizmor: security misconfigurations — unpinned actions, dangerous
+#     `${{ }}` interpolation, missing per-job permissions, etc.
 #
-# Scoped to PRs that touch .github/** only — keeps off the typical PR critical path.
+# Scoped to PRs that touch .github/** only — keeps off the typical PR
+# critical path.
 
 on:
   pull_request:
@@ -17,6 +20,27 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
+  actionlint:
+    name: actionlint
+    runs-on: ubuntu-latest
+    timeout-minutes: 5
+    permissions:
+      contents: read
+    steps:
+      - name: Checkout
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          persist-credentials: false
+
+      # Pinned to v2.1.2. Verify SHA via:
+      #   gh api repos/raven-actions/actionlint/git/refs/tags/v2.1.2
+      # The action wraps the upstream `rhysd/actionlint` binary and emits
+      # GitHub-annotation-formatted findings on PRs.
+      - name: Run actionlint
+        uses: raven-actions/actionlint@205b530c5d9fa8f44ae9ed59f341a0db994aa6f8 # v2.1.2
+        with:
+          fail-on-error: true
+
   zizmor:
     runs-on: ubuntu-latest
     timeout-minutes: 10
diff --git a/.github/zizmor.yml b/.github/zizmor.yml
index c534679c1..b2f89e3ba 100644
--- a/.github/zizmor.yml
+++ b/.github/zizmor.yml
@@ -14,6 +14,15 @@ rules:
       # no checkout of fork code occurs. Header comment in the file documents.
       - ci-report.yml
 
+      # workflow_run is the trusted half of the autofix pipeline. The
+      # untrusted half (pr-autofix.yml) runs fork code with permissions:{}
+      # and produces only a diff artifact (data, not executable code). The
+      # publish job consumes the artifact, allowlist-validates every field
+      # of metadata.json before exporting to $GITHUB_OUTPUT, never checks
+      # out fork code, and never executes anything fork-controlled. Header
+      # comment in the file documents the split.
+      - pr-autofix-publish.yml
+
       # pull_request_target needed by claude-code-action to access secrets
       # and post review comments on fork PRs. Mitigated by: PR checkouts pin
       # the fork's HEAD SHA (not the branch ref) to prevent TOCTOU races,
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index 0ed2aeba5..99930cf0a 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -103,6 +103,24 @@ Every workflow under `.github/workflows/` MUST declare a top-level `concurrency:
 
 - When adding a new workflow, copy the concurrency block from an existing workflow of the same event shape.
 
+## CI automation contracts
+
+Two workflows produce machine-readable signals on every PR. Coding agents and humans alike can rely on the names and shapes below — change them with intent.
+
+### `gitnexus/autofix`
+
+`pr-autofix.yml` (untrusted) + `pr-autofix-publish.yml` (trusted) run `prettier --write` and `eslint --fix` against the PR head and surface the diff as inline review-comment suggestions. Three signals are emitted:
+
+| Surface | Where | Notes |
+|---|---|---|
+| Sticky PR comment | Top-level comment with the HTML marker `<!-- gitnexus:pr-autofix-summary -->` and heading `## :sparkles: PR Autofix`. Only posted when there is something to fix; clean PRs stay silent. | Edit-in-place via marker; one comment per PR. |
+| Fenced JSON block | Inside the sticky, fenced as `gitnexus-autofix`. Schema `gitnexus.pr-autofix/v1` with fields `state` (`suggestions-posted` \| `skipped-too-large`), `pr_number`, `head_sha`, `changed_lines`, `run_id`. | Parseable signal — preferred over regexing prose. |
+| Check Run | Stable name `gitnexus/autofix` on the PR head SHA. Conclusion: `success` (clean) or `neutral` (suggestions-posted / skipped-too-large). The output title disambiguates the two `neutral` cases. | Surfaced under PR Checks; readable via `gh pr checks <pr>`. |
+
+To detect outcome from an agent: `gh pr checks <pr> --json name,conclusion,output | jq '.[] | select(.name == "gitnexus/autofix")'`.
+
+Forks are supported. The untrusted half runs fork code with `permissions: {}` and ships the diff as an artifact; the trusted publish job consumes only the diff (data, not code) and posts the comment + check run.
+
 ## AI-assisted contributions
 
 If you use coding agents, follow project context files (e.g. `AGENTS.md`, `CLAUDE.md`) and avoid drive-by refactors unrelated to the issue. Prefer incremental, test-backed changes.
diff --git a/README.md b/README.md
index 5b7a442b0..0cdb017f4 100644
--- a/README.md
+++ b/README.md
@@ -214,6 +214,7 @@ gitnexus clean --all --force     # Delete all indexes
 gitnexus wiki [path]             # Generate repository wiki from knowledge graph
 gitnexus wiki --model <model>    # Wiki with custom LLM model (default: gpt-4o-mini)
 gitnexus wiki --base-url <url>   # Wiki with custom LLM API base URL
+gitnexus publish                 # Notify the understand-quickly registry (opt-in, see below)
 
 # Repository groups (multi-repo / monorepo service tracking)
 gitnexus group create <name>                                   # Create a repository group
@@ -228,6 +229,12 @@ gitnexus group status <name>     # Check staleness of repos in a group
 
 If `analyze` reports a worker parse timeout on a large or unusual repository, it keeps running and falls back safely. To give slow worker jobs more time, use `gitnexus analyze --worker-timeout 60` or set `GITNEXUS_WORKER_SUB_BATCH_TIMEOUT_MS=60000`. For very large files, `GITNEXUS_WORKER_SUB_BATCH_MAX_BYTES` controls the worker job byte budget.
 
+#### Publishing to understand-quickly (opt-in)
+
+[`looptech-ai/understand-quickly`](https://github.com/looptech-ai/understand-quickly) is a public registry of code-knowledge graphs that lists `gitnexus@1` as a first-class format. After registering your repo once (`npx @understand-quickly/cli add` or the [wizard](https://looptech-ai.github.io/understand-quickly/add.html)), `gitnexus publish` fires a single `repository_dispatch` event so the registry resyncs your entry on demand instead of waiting for the nightly job.
+
+It is opt-in and a no-op without `UNDERSTAND_QUICKLY_TOKEN` — a fine-grained GitHub PAT with `Repository dispatches: write` on the registry repo. Nothing else happens; no graph file is uploaded. See the [protocol spec](https://github.com/looptech-ai/understand-quickly/blob/main/docs/integrations/protocol.md) for the full contract.
+
 ### What Your AI Agent Gets
 
 **16 tools** exposed via MCP (11 per-repo + 5 group):
diff --git a/gitnexus-shared/src/index.ts b/gitnexus-shared/src/index.ts
index ea66c3855..faf136fe7 100644
--- a/gitnexus-shared/src/index.ts
+++ b/gitnexus-shared/src/index.ts
@@ -143,6 +143,18 @@ export type { ScopeTree } from './scope-resolution/scope-tree.js';
 export { buildPositionIndex } from './scope-resolution/position-index.js';
 export type { PositionIndex } from './scope-resolution/position-index.js';
 
+// Understand-Quickly registry integration (opt-in)
+export {
+  UNDERSTAND_QUICKLY_DISPATCH_URL,
+  UNDERSTAND_QUICKLY_EVENT_TYPE,
+  UNDERSTAND_QUICKLY_TOKEN_ENV,
+  buildUqDispatchPayload,
+  isValidOwnerRepo,
+  parseOwnerRepoFromRemote,
+  stripGitSuffix,
+} from './integrations/understand-quickly.js';
+export type { UqDispatchPayload } from './integrations/understand-quickly.js';
+
 // Shadow-mode diff + aggregation (RFC §6.3; Ring 2 SHARED #918)
 export { diffResolutions } from './scope-resolution/shadow/diff.js';
 export type {
diff --git a/gitnexus-shared/src/integrations/understand-quickly.ts b/gitnexus-shared/src/integrations/understand-quickly.ts
new file mode 100644
index 000000000..f30e7461b
--- /dev/null
+++ b/gitnexus-shared/src/integrations/understand-quickly.ts
@@ -0,0 +1,151 @@
+/**
+ * Understand-Quickly registry integration helpers.
+ *
+ * Pure, runtime-agnostic logic for opting in to publishing a GitNexus
+ * index to the [`looptech-ai/understand-quickly`](https://github.com/looptech-ai/understand-quickly)
+ * registry. Lives in `gitnexus-shared` so both the Node CLI and any
+ * future browser-side surface can construct identical dispatch payloads.
+ *
+ * Network I/O lives in the CLI command (`gitnexus/src/cli/publish.ts`)
+ * to keep this module free of Node-only imports — see the comment at
+ * the top of `gitnexus-shared/src/graph/types.ts`.
+ *
+ * The protocol contract (single dispatch event, no graph upload) is
+ * documented at:
+ *   https://github.com/looptech-ai/understand-quickly/blob/main/docs/integrations/protocol.md
+ */
+
+/**
+ * URL of the registry repo's repository_dispatch endpoint. Hardcoded
+ * because the registry is the canonical home for this integration —
+ * users who want a private registry can fork and patch.
+ */
+export const UNDERSTAND_QUICKLY_DISPATCH_URL =
+  'https://api.github.com/repos/looptech-ai/understand-quickly/dispatches';
+
+/**
+ * Event type the registry's sync workflow listens for.
+ * See `looptech-ai/understand-quickly/.github/workflows/sync.yml`.
+ */
+export const UNDERSTAND_QUICKLY_EVENT_TYPE = 'sync-entry';
+
+/** Environment variable that gates the dispatch. */
+export const UNDERSTAND_QUICKLY_TOKEN_ENV = 'UNDERSTAND_QUICKLY_TOKEN';
+
+export interface UqDispatchPayload {
+  event_type: typeof UNDERSTAND_QUICKLY_EVENT_TYPE;
+  client_payload: {
+    /** `<owner>/<repo>` shape — must match the registered entry. */
+    id: string;
+  };
+}
+
+/**
+ * Build the JSON body for the `repository_dispatch` ping. Pure — no
+ * env reads, no network. Validates that `id` looks like `owner/repo`
+ * (one slash, no whitespace, both halves non-empty) so a misconfigured
+ * caller fails loudly before the round-trip.
+ */
+export function buildUqDispatchPayload(id: string): UqDispatchPayload {
+  if (!isValidOwnerRepo(id)) {
+    throw new Error(
+      `[understand-quickly] expected id of the form "owner/repo", got "${id}". ` +
+        `The registry uses this string to look up your entry in registry.json — ` +
+        `it must match the GitHub owner/repo of the source code, not a local path.`,
+    );
+  }
+  return {
+    event_type: UNDERSTAND_QUICKLY_EVENT_TYPE,
+    client_payload: { id },
+  };
+}
+
+/**
+ * `owner/repo` validation. Conservative on purpose: GitHub's actual
+ * naming rules are looser, but we want to catch local paths
+ * (`/Users/...`), bare slugs (`my-repo`), and accidental whitespace.
+ *
+ * Matches GitHub's published slug rules:
+ *   owner: starts with alnum, then alnum/hyphen only, must end with
+ *          alnum (no trailing hyphen — GitHub rejects this at account
+ *          creation, so a `my-org-/repo` input would otherwise pass us
+ *          and 422 from GitHub). No underscore, no dot. Length cap 39.
+ *   repo:  any of alnum/dot/hyphen/underscore. Length cap 100.
+ */
+export function isValidOwnerRepo(id: string): boolean {
+  return /^[A-Za-z0-9](?:[A-Za-z0-9-]{0,37}[A-Za-z0-9])?\/[A-Za-z0-9._-]{1,100}$/.test(id);
+}
+
+/**
+ * Strip a single trailing `.git` (case-insensitive) and any trailing
+ * slashes from a URL-ish string. Bounded linear: each character is
+ * visited at most twice, no backtracking.
+ *
+ * Replaces `s.replace(/\.git\/*$/i, '').replace(/\/+$/, '')` which
+ * CodeQL's polynomial-regex check (codeql/js/polynomial-redos) flags as
+ * a worst-case O(n²) on adversarial input like "////.../x".
+ */
+export function stripGitSuffix(input: string): string {
+  let end = input.length;
+  // Trim trailing '/'.
+  while (end > 0 && input.charCodeAt(end - 1) === 0x2f) end--;
+  // Drop one trailing '.git' (case-insensitive).
+  if (end >= 4) {
+    const tail = input.slice(end - 4, end).toLowerCase();
+    if (tail === '.git') end -= 4;
+  }
+  // Trim trailing '/' that may have sat between '.git' and the rest.
+  while (end > 0 && input.charCodeAt(end - 1) === 0x2f) end--;
+  return input.slice(0, end);
+}
+
+/**
+ * Parse `owner/repo` out of a git remote URL. Mirrors the heuristic in
+ * `gitnexus/src/storage/git.ts:parseRepoNameFromUrl` but keeps both
+ * halves so we can build a registry id. Returns `null` on shapes we
+ * don't recognise.
+ *
+ * Examples:
+ *   git@github.com:looptech-ai/understand-quickly.git
+ *   https://github.com/looptech-ai/understand-quickly
+ *   ssh://git@github.com/looptech-ai/understand-quickly.git
+ */
+export function parseOwnerRepoFromRemote(url: string | null | undefined): string | null {
+  if (!url) return null;
+  const trimmed = url.trim();
+  if (!trimmed) return null;
+  // Strip a trailing `.git` (case-insensitive) and any trailing slashes
+  // so https://h/o/r and https://h/o/r.git collapse to the same id.
+  // Bounded-linear helper avoids the polynomial-regex CodeQL alert.
+  const stripped = stripGitSuffix(trimmed);
+
+  // SCP-form SSH (`git@host:owner/repo`). Capture host so we can reject
+  // non-GitHub remotes — a GitLab origin like
+  // `https://gitlab.example.com/group/sub/project.git` would otherwise
+  // silently dispatch the wrong id (LOW 9).
+  const ssh = stripped.match(/^[^@]+@([^:]+):([^/]+)\/([^/]+)$/);
+  if (ssh) {
+    const host = ssh[1].toLowerCase();
+    if (host !== 'github.com' && host !== 'www.github.com') return null;
+    return `${ssh[2]}/${ssh[3]}`;
+  }
+
+  // URL forms (https://, ssh://, git://, file://) — last two path segments.
+  const url2 = stripped.match(/^[a-zA-Z][a-zA-Z0-9+.-]*:\/\/([^/]+)\/(.+)$/);
+  if (url2) {
+    // Strip optional `userinfo@` (e.g. `ssh://git@github.com/...`).
+    const authority = url2[1];
+    const atIdx = authority.lastIndexOf('@');
+    const hostAndPort = atIdx >= 0 ? authority.slice(atIdx + 1) : authority;
+    // Strip `:port` suffix if present.
+    const colonIdx = hostAndPort.indexOf(':');
+    const host = (colonIdx >= 0 ? hostAndPort.slice(0, colonIdx) : hostAndPort).toLowerCase();
+    if (host !== 'github.com' && host !== 'www.github.com') return null;
+    const segments = url2[2].split('/').filter(Boolean);
+    if (segments.length >= 2) {
+      const [owner, repo] = segments.slice(-2);
+      return `${owner}/${repo}`;
+    }
+  }
+  return null;
+}
diff --git a/gitnexus/package-lock.json b/gitnexus/package-lock.json
index b07ccd7cb..04233b1a1 100644
--- a/gitnexus/package-lock.json
+++ b/gitnexus/package-lock.json
@@ -2065,9 +2065,9 @@
       "license": "MIT"
     },
     "node_modules/@types/node": {
-      "version": "25.6.1",
-      "resolved": "https://registry.npmjs.org/@types/node/-/node-25.6.1.tgz",
-      "integrity": "sha512-coJCN8O1q4AGyyqCAUSP06P+SrMTu18BkEj3NVAK07q6QUneD2wzj3CLv9+yP+BMeZQlMvneXqqvDe3w+xcq7g==",
+      "version": "25.6.2",
+      "resolved": "https://registry.npmjs.org/@types/node/-/node-25.6.2.tgz",
+      "integrity": "sha512-sokuT28dxf9JT5Kady1fsXOvI4HVpjZa95NKT5y9PNTIrs2AsobR4GFAA90ZG8M+nxVRLysCXsVj6eGC7Vbrlw==",
       "license": "MIT",
       "dependencies": {
         "undici-types": "~7.19.0"
@@ -3110,9 +3110,9 @@
       "license": "MIT"
     },
     "node_modules/fast-uri": {
-      "version": "3.1.0",
-      "resolved": "https://registry.npmjs.org/fast-uri/-/fast-uri-3.1.0.tgz",
-      "integrity": "sha512-iPeeDKJSWf4IEOasVVrknXpaBV0IApz/gp7S2bb7Z4Lljbl2MGJRqInZiUrQwV16cpzw/D3S5j5Julj/gT52AA==",
+      "version": "3.1.2",
+      "resolved": "https://registry.npmjs.org/fast-uri/-/fast-uri-3.1.2.tgz",
+      "integrity": "sha512-rVjf7ArG3LTk+FS6Yw81V1DLuZl1bRbNrev6Tmd/9RaroeeRRJhAt7jg/6YFxbvAQXUCavSoZhPPj6oOx+5KjQ==",
       "funding": [
         {
           "type": "github",
@@ -3476,9 +3476,9 @@
       "license": "MIT"
     },
     "node_modules/hono": {
-      "version": "4.12.16",
-      "resolved": "https://registry.npmjs.org/hono/-/hono-4.12.16.tgz",
-      "integrity": "sha512-jN0ZewiNAWSe5khM3EyCmBb250+b40wWbwNILNfEvq84VREWwOIkuUsFONk/3i3nqkz7Oe1PcpM2mwQEK2L9Kg==",
+      "version": "4.12.18",
+      "resolved": "https://registry.npmjs.org/hono/-/hono-4.12.18.tgz",
+      "integrity": "sha512-RWzP96k/yv0PQfyXnWjs6zot20TqfpfsNXhOnev8d1InAxubW93L11/oNUc3tQqn2G0bSdAOBpX+2uDFHV7kdQ==",
       "license": "MIT",
       "engines": {
         "node": ">=16.9.0"
@@ -4282,15 +4282,15 @@
       }
     },
     "node_modules/onnxruntime-common": {
-      "version": "1.25.1",
-      "resolved": "https://registry.npmjs.org/onnxruntime-common/-/onnxruntime-common-1.25.1.tgz",
-      "integrity": "sha512-kKvYQFdos4LWJqhZ+nmKu3NT8NXzw8I5x9fNUKe1rNKcPfNKnYXUtW7JBpcKFsvLtrJashRgVYSbFap4cHxvNg==",
+      "version": "1.26.0",
+      "resolved": "https://registry.npmjs.org/onnxruntime-common/-/onnxruntime-common-1.26.0.tgz",
+      "integrity": "sha512-qVyMR4lcWgbkc4getFV+GQijsTnbg/siteoqcDwa3sI/LxbrMSNw4ePyvCq/ymdQaRomCA7YuWmhzsswxvymdw==",
       "license": "MIT"
     },
     "node_modules/onnxruntime-node": {
-      "version": "1.25.1",
-      "resolved": "https://registry.npmjs.org/onnxruntime-node/-/onnxruntime-node-1.25.1.tgz",
-      "integrity": "sha512-N0M58CGTiTsLkPpx9bxmRFi24GT6r67Qei/GrBEIiDyntcYdXU5vQZp112ypydG9vEKRFgbgUYQJnEi+jll8dg==",
+      "version": "1.26.0",
+      "resolved": "https://registry.npmjs.org/onnxruntime-node/-/onnxruntime-node-1.26.0.tgz",
+      "integrity": "sha512-OHl6PiOEOqxaLHL0N9eFrbzS7IGmu3BtJNH3RTEnRAheCIkfc3gjcjl4sGcjp9C22ZC9YTquDOxSdT/stBQ6BQ==",
       "hasInstallScript": true,
       "license": "MIT",
       "os": [
@@ -4301,7 +4301,7 @@
       "dependencies": {
         "adm-zip": "^0.5.16",
         "global-agent": "^4.1.3",
-        "onnxruntime-common": "1.25.1"
+        "onnxruntime-common": "1.26.0"
       }
     },
     "node_modules/onnxruntime-web": {
diff --git a/gitnexus/src/cli/analyze.ts b/gitnexus/src/cli/analyze.ts
index 175db47e6..47745c575 100644
--- a/gitnexus/src/cli/analyze.ts
+++ b/gitnexus/src/cli/analyze.ts
@@ -27,6 +27,7 @@ import { warnMissingOptionalGrammars } from './optional-grammars.js';
 import { glob } from 'glob';
 import fs from 'fs/promises';
 import { cliError } from './cli-message.js';
+import { isHfDownloadFailure } from '../core/embeddings/hf-env.js';
 
 // Capture stderr.write at module load BEFORE anything (LadybugDB native
 // init, progress bar, console redirection) can monkey-patch it. The
@@ -576,6 +577,26 @@ export const analyzeCommand = async (inputPath?: string, options?: AnalyzeOption
       return;
     }
 
+    // HF download failure — show clean guidance without the raw stack trace.
+    // Checked before writeFatalToStderr so the user sees one focused message
+    // rather than a stack-trace dump followed by a second remediation block.
+    if (isHfDownloadFailure(msg) || msg.includes('Failed to download embedding model')) {
+      cliError(
+        `  The embedding model could not be downloaded.\n` +
+          `  huggingface.co may be unreachable from your network\n` +
+          `  (e.g. behind a corporate proxy or a regional firewall).\n` +
+          `  Suggestions:\n` +
+          `    1. Set HF_ENDPOINT to a mirror and retry:\n` +
+          `         HF_ENDPOINT=https://hf-mirror.com npx gitnexus analyze --embeddings\n` +
+          `         (Windows: set HF_ENDPOINT=https://hf-mirror.com && npx gitnexus analyze --embeddings)\n` +
+          `    2. Check your proxy / VPN settings.\n` +
+          `    3. Once downloaded the model is cached — future runs work offline.\n`,
+        { recoveryHint: 'hf-endpoint-unreachable' },
+      );
+      process.exitCode = 1;
+      return;
+    }
+
     // Bypass the redirected console.error and write the full stack to
     // the real stderr captured at module load. The redirected
     // console.error wraps every line with `\\x1b[2K\\r` (ANSI clear-line)
diff --git a/gitnexus/src/cli/index.ts b/gitnexus/src/cli/index.ts
index bde0a25d6..5fb0e2384 100644
--- a/gitnexus/src/cli/index.ts
+++ b/gitnexus/src/cli/index.ts
@@ -163,6 +163,18 @@ program
   .description('Augment a search pattern with knowledge graph context (used by hooks)')
   .action(createLazyAction(() => import('./augment.js'), 'augmentCommand'));
 
+program
+  .command('publish [path]')
+  .description(
+    'Notify the understand-quickly registry that this repo has a fresh GitNexus index. ' +
+      'Opt-in: requires UNDERSTAND_QUICKLY_TOKEN (fine-grained PAT with ' +
+      '`Repository dispatches: write` on looptech-ai/understand-quickly). ' +
+      'No-op without the token. See https://github.com/looptech-ai/understand-quickly.',
+  )
+  .option('--id <owner/repo>', 'Override the registry id (defaults to the origin remote)')
+  .option('--skip-git', 'Treat cwd as the repo root and skip parent git-root discovery')
+  .action(createLazyAction(() => import('./publish.js'), 'publishCommand'));
+
 // ─── Direct Tool Commands (no MCP overhead) ────────────────────────
 // These invoke LocalBackend directly for use in eval, scripts, and CI.
 
diff --git a/gitnexus/src/cli/publish.ts b/gitnexus/src/cli/publish.ts
new file mode 100644
index 000000000..8aedc9c35
--- /dev/null
+++ b/gitnexus/src/cli/publish.ts
@@ -0,0 +1,232 @@
+/**
+ * `gitnexus publish` — opt-in ping to the understand-quickly registry.
+ *
+ * Fires a single `repository_dispatch` event at
+ * `looptech-ai/understand-quickly` so the registry knows to refresh its
+ * entry for the current repo. Does NOT upload anything: per the
+ * understand-quickly protocol, the registry pulls the graph from a
+ * raw-GitHub URL the user controls.
+ *
+ *   https://github.com/looptech-ai/understand-quickly/blob/main/docs/integrations/protocol.md
+ *
+ * Defaults:
+ *   - Without `UNDERSTAND_QUICKLY_TOKEN` in the env, this is a no-op
+ *     (prints one informational line, exit 0). Same shape as the
+ *     `--publish` patterns in sibling tools.
+ *   - With the token, fires the dispatch and reports the response code.
+ *
+ * The `id` is derived from the repo's `origin` remote unless the caller
+ * passes `--id <owner/repo>` explicitly. We deliberately do NOT auto-add
+ * the repo to the registry — registration is one-time and uses the
+ * `npx @understand-quickly/cli add` path documented in the protocol.
+ */
+
+import path from 'path';
+import {
+  UNDERSTAND_QUICKLY_DISPATCH_URL,
+  UNDERSTAND_QUICKLY_TOKEN_ENV,
+  buildUqDispatchPayload,
+  isValidOwnerRepo,
+  parseOwnerRepoFromRemote,
+} from 'gitnexus-shared';
+import { getGitRoot, getRemoteOriginUrl, getCurrentCommit } from '../storage/git.js';
+import { hasIndex } from '../storage/repo-manager.js';
+import { cliInfo, cliError } from './cli-message.js';
+
+export interface PublishOptions {
+  /** Override the auto-derived `owner/repo` id. */
+  id?: string;
+  /** Treat the cwd as the repo root (skip git-root walk). */
+  skipGit?: boolean;
+}
+
+const REGISTER_HINT =
+  'Register your repo once with: npx @understand-quickly/cli add\n' +
+  'Or use the wizard: https://looptech-ai.github.io/understand-quickly/add.html';
+
+/**
+ * Hard cap on the dispatch fetch to keep CI publish steps from stalling
+ * for the OS TCP timeout (~2 min) when api.github.com is unreachable.
+ * Matches the pattern used in `src/core/embeddings/http-client.ts`.
+ */
+const DISPATCH_TIMEOUT_MS = 15_000;
+
+export const publishCommand = async (
+  inputPath?: string,
+  options: PublishOptions = {},
+): Promise<void> => {
+  // ── 0. Token gate FIRST — guarantees true no-op without the token. ──
+  // The README, CLI --help, and PR body all promise "exit 0 without
+  // UNDERSTAND_QUICKLY_TOKEN". Doing the index/repo-root checks before
+  // the token gate would make those promises false for users who haven't
+  // run `gitnexus analyze` yet but want to verify the command is wired.
+  const token = process.env[UNDERSTAND_QUICKLY_TOKEN_ENV];
+  if (!token) {
+    cliInfo(
+      `[understand-quickly] ${UNDERSTAND_QUICKLY_TOKEN_ENV} is not set — skipping dispatch.\n` +
+        `Set it to a fine-grained PAT with "Repository dispatches: write" on ` +
+        `looptech-ai/understand-quickly to enable instant resync.\n` +
+        `(Without the token, the registry's nightly sync still picks up your entry.)`,
+      { skipped: 'no-token' },
+    );
+    return;
+  }
+
+  // ── 1. Resolve the repo root (same precedence as `analyze`) ──────────
+  let repoPath: string;
+  if (inputPath) {
+    repoPath = path.resolve(inputPath);
+  } else if (options.skipGit) {
+    repoPath = path.resolve(process.cwd());
+  } else {
+    const gitRoot = getGitRoot(process.cwd());
+    if (!gitRoot) {
+      cliError(
+        '[understand-quickly] not inside a git repository.\n' +
+          'Run from a repo, or pass --skip-git to publish from the current directory.',
+      );
+      process.exitCode = 1;
+      return;
+    }
+    repoPath = gitRoot;
+  }
+
+  // ── 2. Confirm a GitNexus index exists ───────────────────────────────
+  // Publishing without an index is almost always a mistake — the
+  // registry's nightly sync would fetch a stale or missing graph file
+  // and mark the entry `missing`. Refuse loudly with a fix-it hint.
+  if (!(await hasIndex(repoPath))) {
+    cliError(
+      `[understand-quickly] no GitNexus index found at ${repoPath}/.gitnexus.\n` +
+        'Run `gitnexus analyze` first, then re-run `gitnexus publish`.',
+    );
+    process.exitCode = 1;
+    return;
+  }
+
+  // ── 3. Derive the registry id ─────────────────────────────────────────
+  const id =
+    options.id ?? parseOwnerRepoFromRemote(getRemoteOriginUrl(repoPath) ?? undefined) ?? null;
+  if (!id || !isValidOwnerRepo(id)) {
+    cliError(
+      `[understand-quickly] could not derive a registry id from this repo.\n` +
+        `Pass --id <owner/repo> explicitly (e.g. --id looptech-ai/${path.basename(repoPath)}).\n` +
+        REGISTER_HINT,
+    );
+    process.exitCode = 1;
+    return;
+  }
+
+  // ── 4. Fire the dispatch ─────────────────────────────────────────────
+  const payload = buildUqDispatchPayload(id);
+  let response: Response;
+  try {
+    response = await fetch(UNDERSTAND_QUICKLY_DISPATCH_URL, {
+      method: 'POST',
+      headers: {
+        Accept: 'application/vnd.github+json',
+        Authorization: `Bearer ${token}`,
+        'X-GitHub-Api-Version': '2022-11-28',
+        'Content-Type': 'application/json',
+        'User-Agent': 'gitnexus-cli',
+      },
+      body: JSON.stringify(payload),
+      signal: AbortSignal.timeout(DISPATCH_TIMEOUT_MS),
+    });
+  } catch (err) {
+    // `AbortSignal.timeout()` throws a `DOMException` with `name ===
+    // 'TimeoutError'` on Node 18.14+ (and on browsers/Bun). It is NOT
+    // a plain `AbortError`. Match the pattern used in
+    // gitnexus/src/core/embeddings/http-client.ts so the user sees the
+    // targeted "timed out" message instead of a generic "operation
+    // was aborted".
+    const isTimeout = err instanceof DOMException && err.name === 'TimeoutError';
+    if (isTimeout) {
+      cliError(
+        `[understand-quickly] dispatch timed out after ${DISPATCH_TIMEOUT_MS}ms. ` +
+          `Check network access to api.github.com and retry.`,
+        { id },
+      );
+    } else {
+      const msg = err instanceof Error ? err.message : String(err);
+      cliError(`[understand-quickly] dispatch network error: ${msg}`, { id });
+    }
+    process.exitCode = 1;
+    return;
+  }
+
+  // GitHub returns 204 on success. Distinct branches for 401/403/404/422
+  // so users debug without checking the docs.
+  if (response.status === 204) {
+    await response.body?.cancel().catch(() => {});
+    // `getCurrentCommit` is only meaningful in the success path — moving
+    // it inside this branch removes a wasted child-process spawn on every
+    // error response (LOW 7).
+    const commit = getCurrentCommit(repoPath);
+    cliInfo(
+      `[understand-quickly] dispatched sync-entry for ${id}` +
+        (commit ? ` @ ${commit.slice(0, 7)}` : '') +
+        '.\n' +
+        `Note: a 204 only confirms GitHub accepted the dispatch. Whether the ` +
+        `registry workflow finds an entry for "${id}" is logged at ` +
+        `https://github.com/looptech-ai/understand-quickly/actions/workflows/sync.yml`,
+      { id, commit, status: response.status },
+    );
+    return;
+  }
+
+  if (response.status === 401) {
+    cliError(
+      `[understand-quickly] dispatch returned 401 — the ${UNDERSTAND_QUICKLY_TOKEN_ENV} value is invalid or expired.\n` +
+        `Regenerate a fine-grained PAT at https://github.com/settings/personal-access-tokens ` +
+        `with Repository access scoped to looptech-ai/understand-quickly and the ` +
+        `"Repository dispatches: write" permission, then retry.`,
+      { id, status: response.status },
+    );
+    process.exitCode = 1;
+    return;
+  }
+
+  if (response.status === 403) {
+    cliError(
+      `[understand-quickly] dispatch returned 403 — the token authenticated but ` +
+        `lacks the "Repository dispatches: write" permission on ` +
+        `looptech-ai/understand-quickly. Edit the PAT scopes and retry.`,
+      { id, status: response.status },
+    );
+    process.exitCode = 1;
+    return;
+  }
+
+  if (response.status === 404) {
+    cliError(
+      `[understand-quickly] dispatch returned 404 — the token cannot reach ` +
+        `looptech-ai/understand-quickly. Verify the PAT has Repository access to ` +
+        `that exact repo (not just your own org).`,
+      { id, status: response.status },
+    );
+    process.exitCode = 1;
+    return;
+  }
+
+  if (response.status === 422) {
+    // Malformed event_type / client_payload — a code bug in this CLI,
+    // not a user mistake. Surface so we get bug reports.
+    const body422 = await response.text().catch(() => '');
+    cliError(
+      `[understand-quickly] dispatch returned 422 (this is a CLI bug; please report).\n` +
+        `Body: ${body422 || '(empty)'}`,
+      { id, status: response.status },
+    );
+    process.exitCode = 1;
+    return;
+  }
+
+  // 5xx and anything else → bubble the body so the user has something to act on.
+  const body = await response.text().catch(() => '');
+  cliError(
+    `[understand-quickly] dispatch failed with HTTP ${response.status}: ${body || '(empty body)'}`,
+    { id, status: response.status },
+  );
+  process.exitCode = 1;
+};
diff --git a/gitnexus/src/config/ignore-service.ts b/gitnexus/src/config/ignore-service.ts
index ce1fda913..2c3eebe3b 100644
--- a/gitnexus/src/config/ignore-service.ts
+++ b/gitnexus/src/config/ignore-service.ts
@@ -25,6 +25,8 @@ const DEFAULT_IGNORE_LIST = new Set([
   'bower_components',
   'jspm_packages',
   'vendor', // PHP/Go
+  'third_party', // C/C++ (Google-style vendored dependencies)
+  '3rdparty', // C/C++ (alternate spelling, also Qt convention)
   // 'packages' removed - commonly used for monorepo source code (lerna, pnpm, yarn workspaces)
   'venv',
   '.venv',
diff --git a/gitnexus/src/core/augmentation/engine.ts b/gitnexus/src/core/augmentation/engine.ts
index 896813c22..f97415cc9 100644
--- a/gitnexus/src/core/augmentation/engine.ts
+++ b/gitnexus/src/core/augmentation/engine.ts
@@ -104,7 +104,7 @@ export async function augment(pattern: string, cwd?: string): Promise<string> {
     }
 
     // Step 1: BM25 search (fast, no embeddings)
-    const bm25Results = await searchFTSFromLbug(pattern, 10, repoId);
+    const { results: bm25Results } = await searchFTSFromLbug(pattern, 10, repoId);
 
     if (bm25Results.length === 0) return '';
 
diff --git a/gitnexus/src/core/embeddings/embedder.ts b/gitnexus/src/core/embeddings/embedder.ts
index b37fb45f3..72ddcbd70 100644
--- a/gitnexus/src/core/embeddings/embedder.ts
+++ b/gitnexus/src/core/embeddings/embedder.ts
@@ -14,7 +14,12 @@ if (!process.env.ORT_LOG_LEVEL) {
   process.env.ORT_LOG_LEVEL = '3';
 }
 
-import { pipeline, env, type FeatureExtractionPipeline } from '@huggingface/transformers';
+import {
+  pipeline,
+  env,
+  type FeatureExtractionPipeline,
+  type ProgressInfo,
+} from '@huggingface/transformers';
 import { existsSync } from 'fs';
 import { execFileSync } from 'child_process';
 import { join, dirname } from 'path';
@@ -22,7 +27,7 @@ import { createRequire } from 'module';
 import { DEFAULT_EMBEDDING_CONFIG, type EmbeddingConfig, type ModelProgress } from './types.js';
 import { isHttpMode, getHttpDimensions, httpEmbed } from './http-client.js';
 import { resolveEmbeddingConfig } from './config.js';
-import { applyHfEnvOverrides } from './hf-env.js';
+import { applyHfEnvOverrides, isHfDownloadFailure, withHfDownloadRetry } from './hf-env.js';
 import { logger } from '../logger.js';
 
 /**
@@ -171,13 +176,18 @@ export const initEmbedder = async (
       }
 
       const progressCallback = onProgress
-        ? (data: any) => {
+        ? (data: ProgressInfo) => {
             const progress: ModelProgress = {
-              status: data.status || 'progress',
-              file: data.file,
-              progress: data.progress,
-              loaded: data.loaded,
-              total: data.total,
+              // Map the `progress_total` aggregate event (not in ModelProgress.status)
+              // back to 'progress' so callers don't need to handle it separately.
+              status:
+                data.status === 'progress_total'
+                  ? 'progress'
+                  : ((data.status as ModelProgress['status']) ?? 'progress'),
+              file: 'file' in data ? data.file : undefined,
+              progress: 'progress' in data ? data.progress : undefined,
+              loaded: 'loaded' in data ? data.loaded : undefined,
+              total: 'total' in data ? data.total : undefined,
             };
             onProgress(progress);
           }
@@ -202,17 +212,29 @@ export const initEmbedder = async (
             logger.info('🔧 Using WASM backend (slower)...');
           }
 
-          embedderInstance = await (pipeline as any)('feature-extraction', finalConfig.modelId, {
-            device: device,
-            dtype: 'fp32',
-            progress_callback: progressCallback,
-            session_options: {
-              logSeverityLevel: 3,
-              intraOpNumThreads: finalConfig.threads,
-              interOpNumThreads: 1,
-              executionMode: 'sequential',
+          embedderInstance = await withHfDownloadRetry(
+            () =>
+              pipeline('feature-extraction', finalConfig.modelId, {
+                device: device,
+                dtype: 'fp32',
+                progress_callback: progressCallback,
+                session_options: {
+                  logSeverityLevel: 3,
+                  intraOpNumThreads: finalConfig.threads,
+                  interOpNumThreads: 1,
+                  executionMode: 'sequential',
+                },
+              }),
+            {
+              onRetry: isDev
+                ? (attempt, max, err) =>
+                    logger.warn(
+                      { attempt, max, err: err.message },
+                      `⚠️  Model download network error (attempt ${attempt}/${max}), retrying…`,
+                    )
+                : undefined,
             },
-          });
+          );
           currentDevice = device;
 
           if (isDev) {
@@ -228,6 +250,20 @@ export const initEmbedder = async (
 
           return embedderInstance!;
         } catch (deviceError) {
+          // Network errors and circuit-open errors are not device-specific —
+          // they will fail the same way on every device. Rethrow immediately
+          // with actionable HF_ENDPOINT guidance rather than silently falling
+          // back to the next device.
+          const errMsg = deviceError instanceof Error ? deviceError.message : String(deviceError);
+          if (isHfDownloadFailure(errMsg)) {
+            const endpointHint = process.env.HF_ENDPOINT
+              ? `The configured endpoint (${process.env.HF_ENDPOINT}) may be unreachable.`
+              : `huggingface.co may be unreachable from your network.\n` +
+                `  Set HF_ENDPOINT to a mirror and retry:\n` +
+                `    HF_ENDPOINT=https://hf-mirror.com npx gitnexus analyze --embeddings\n` +
+                `    (Windows: set HF_ENDPOINT=https://hf-mirror.com && npx gitnexus analyze --embeddings)`;
+            throw new Error(`Failed to download embedding model: ${errMsg}\n  ${endpointHint}`);
+          }
           if (isDev && (device === 'cuda' || device === 'dml')) {
             const gpuType = device === 'dml' ? 'DirectML' : 'CUDA';
             logger.info(`⚠️  ${gpuType} not available, falling back to CPU...`);
diff --git a/gitnexus/src/core/embeddings/hf-env.ts b/gitnexus/src/core/embeddings/hf-env.ts
index 6a977a76d..95548fff2 100644
--- a/gitnexus/src/core/embeddings/hf-env.ts
+++ b/gitnexus/src/core/embeddings/hf-env.ts
@@ -1,6 +1,25 @@
 import os from 'node:os';
 import { join } from 'node:path';
 
+// ---------------------------------------------------------------------------
+// Download resilience defaults
+// ---------------------------------------------------------------------------
+
+/** Per-attempt timeout for the full model download (5 minutes). */
+export const HF_DOWNLOAD_TIMEOUT_MS = 5 * 60 * 1_000;
+/** Maximum total download attempts (1 initial + N-1 retries). */
+export const HF_MAX_ATTEMPTS = 3;
+/** Initial delay between retry attempts; doubles on each subsequent retry. */
+export const HF_BASE_DELAY_MS = 2_000;
+/** Number of consecutive failures required to open the circuit. */
+export const CB_FAILURE_THRESHOLD = 3;
+/** How long the circuit stays open before transitioning to half-open. */
+export const CB_RESET_TIMEOUT_MS = 60_000;
+/** Upper bound clamped on the env-override per-attempt timeout (30 minutes). */
+export const HF_MAX_TIMEOUT_MS = 30 * 60 * 1_000;
+/** Upper bound clamped on the env-override attempt count. */
+export const HF_MAX_ATTEMPTS_CAP = 10;
+
 /**
  * @internal Exported only for unit tests and the two embedder entry points
  * (`core/embeddings/embedder.ts` + `mcp/core/embedder.ts`). Not part of the
@@ -60,3 +79,265 @@ export function applyHfEnvOverrides(env: HfEnvSubset): void {
     env.remoteHost = endpoint.endsWith('/') ? endpoint : endpoint + '/';
   }
 }
+
+/**
+ * @internal Exported for unit tests and the two embedder entry points.
+ *
+ * Returns true when an error message indicates a network-level fetch failure
+ * during HuggingFace model download (e.g. `TypeError: fetch failed`,
+ * `ECONNREFUSED`, `ENOTFOUND`, `ETIMEDOUT`, `ECONNRESET`).
+ *
+ * These errors are not device-specific and cannot be fixed by falling back to
+ * a different ONNX device — the caller should rethrow immediately with
+ * guidance about `HF_ENDPOINT`.
+ */
+export function isNetworkFetchError(message: string): boolean {
+  return (
+    message.includes('fetch failed') ||
+    message.includes('ECONNREFUSED') ||
+    message.includes('ENOTFOUND') ||
+    message.includes('ETIMEDOUT') ||
+    message.includes('ECONNRESET')
+  );
+}
+
+// ---------------------------------------------------------------------------
+// Circuit breaker
+// ---------------------------------------------------------------------------
+
+/** @internal Used by `withHfDownloadRetry` to mark a circuit-open rejection. */
+export const CIRCUIT_OPEN_TAG = 'hf-circuit-open';
+
+/** Circuit-breaker states. */
+type CircuitState = 'closed' | 'open' | 'half-open';
+
+/**
+ * Circuit breaker for HuggingFace model downloads.
+ *
+ * After `failureThreshold` consecutive network failures the circuit opens and
+ * all subsequent calls to `withHfDownloadRetry` fail immediately without
+ * issuing any network requests. After `resetTimeoutMs` the circuit enters the
+ * half-open state and the next call is attempted — if it succeeds the circuit
+ * closes again; if it fails the circuit re-opens.
+ *
+ * Exported for unit-testing; production code should use the module-level
+ * `hfDownloadCircuit` singleton.
+ */
+export class HfDownloadCircuitBreaker {
+  private _state: CircuitState = 'closed';
+  private _failures = 0;
+  /** Timestamp of the last recorded failure (ms since epoch). */
+  lastFailureAt = 0;
+
+  constructor(
+    readonly failureThreshold: number = CB_FAILURE_THRESHOLD,
+    readonly resetTimeoutMs: number = CB_RESET_TIMEOUT_MS,
+  ) {}
+
+  /** Effective state, factoring in the reset-timeout transition. */
+  get state(): CircuitState {
+    if (this._state === 'open' && Date.now() - this.lastFailureAt > this.resetTimeoutMs) {
+      this._state = 'half-open';
+    }
+    return this._state;
+  }
+
+  /** Returns true when the circuit is open and calls should be rejected. */
+  isOpen(): boolean {
+    return this.state === 'open';
+  }
+
+  /** Record a successful call — resets the failure counter and closes the circuit. */
+  recordSuccess(): void {
+    this._failures = 0;
+    this._state = 'closed';
+  }
+
+  /** Record a failed call — increments the counter and opens the circuit when the threshold is reached. */
+  recordFailure(): void {
+    this._failures++;
+    this.lastFailureAt = Date.now();
+    if (this._failures >= this.failureThreshold) {
+      this._state = 'open';
+    }
+  }
+
+  /** @internal Reset to initial state (used in tests). */
+  reset(): void {
+    this._failures = 0;
+    this._state = 'closed';
+    this.lastFailureAt = 0;
+  }
+}
+
+/** Module-level singleton shared by both embedder entry points. */
+export const hfDownloadCircuit = new HfDownloadCircuitBreaker();
+
+// ---------------------------------------------------------------------------
+// Retry + timeout wrapper
+// ---------------------------------------------------------------------------
+
+/** @internal Returns true for errors that should abort without retry (circuit-open). */
+export function isHfCircuitOpenError(message: string): boolean {
+  return message.includes(CIRCUIT_OPEN_TAG);
+}
+
+/**
+ * Returns true for any HuggingFace download failure that warrants showing the
+ * `HF_ENDPOINT` remediation hint: either a raw network error or a
+ * circuit-open rejection (which itself was caused by repeated network errors).
+ */
+export function isHfDownloadFailure(message: string): boolean {
+  return isNetworkFetchError(message) || isHfCircuitOpenError(message);
+}
+
+/** @internal Wraps `fn` in a hard time-limit. The timeout error contains
+ *  `ETIMEDOUT` so that `isNetworkFetchError` classifies it correctly.
+ */
+export function withDownloadTimeout<T>(fn: () => Promise<T>, timeoutMs: number): Promise<T> {
+  return new Promise<T>((resolve, reject) => {
+    const timer = setTimeout(
+      () =>
+        reject(
+          new Error(
+            `ETIMEDOUT: model download timed out after ${Math.round(timeoutMs / 1000)}s — ` +
+              `check your network speed or set HF_ENDPOINT to a faster mirror`,
+          ),
+        ),
+      timeoutMs,
+    );
+    fn().then(
+      (v) => {
+        clearTimeout(timer);
+        resolve(v);
+      },
+      (e) => {
+        clearTimeout(timer);
+        reject(e);
+      },
+    );
+  });
+}
+
+/** @internal Async sleep (exposed for testing). */
+export function sleep(ms: number): Promise<void> {
+  return new Promise((resolve) => setTimeout(resolve, ms));
+}
+
+export interface HfRetryOptions {
+  /** Maximum total attempts including the initial one (default: `HF_MAX_ATTEMPTS`). */
+  maxAttempts?: number;
+  /** Delay before the first retry; doubles on each subsequent attempt (default: `HF_BASE_DELAY_MS`). */
+  baseDelayMs?: number;
+  /** Per-attempt wall-clock timeout in ms (default: `HF_DOWNLOAD_TIMEOUT_MS`). */
+  timeoutMs?: number;
+  /**
+   * Circuit-breaker instance to use. Defaults to the module-level
+   * `hfDownloadCircuit` singleton. Pass a fresh instance in tests.
+   */
+  circuit?: HfDownloadCircuitBreaker;
+  /**
+   * Optional callback invoked before each retry (not the initial attempt).
+   * @param attempt - 1-based retry number
+   * @param max - total allowed attempts
+   * @param error - the error that triggered the retry
+   */
+  onRetry?: (attempt: number, max: number, error: Error) => void;
+}
+
+/**
+ * Retry wrapper for HuggingFace model downloads with per-attempt timeout and
+ * circuit-breaker protection.
+ *
+ * Behaviour:
+ * - If the circuit is **open**, fails immediately with a `CIRCUIT_OPEN_TAG`
+ *   message (so `isHfDownloadFailure` still returns true and the caller can
+ *   show `HF_ENDPOINT` guidance).
+ * - Each attempt is wrapped in `withDownloadTimeout`.
+ * - On a network-level error (`isNetworkFetchError`) the attempt is retried
+ *   with exponential back-off; non-network errors (e.g. ONNX device failure)
+ *   are rethrown immediately without retry.
+ * - Every network failure is recorded on the circuit breaker; a success resets
+ *   it.
+ * - After all attempts are exhausted, the last network error is rethrown
+ *   so the existing `isNetworkFetchError` / `isHfDownloadFailure` guards in
+ *   the calling code still fire.
+ */
+export async function withHfDownloadRetry<T>(
+  fn: () => Promise<T>,
+  options: HfRetryOptions = {},
+): Promise<T> {
+  // Resolve effective values — explicit options take precedence over env vars,
+  // which take precedence over built-in defaults. This lets users lower the
+  // per-attempt timeout without rebuilding (e.g.
+  //   HF_DOWNLOAD_TIMEOUT_MS=60000 npx gitnexus analyze --embeddings
+  // reduces the worst-case wait from 15 minutes to ~3 minutes).
+  //
+  // Upper bounds are clamped to prevent accidental runaway configuration:
+  //   - timeoutMs is capped at HF_MAX_TIMEOUT_MS (30 min)
+  //   - maxAttempts is floored (fractional values → integer) and capped at
+  //     HF_MAX_ATTEMPTS_CAP (10).  Values ≤ 0, NaN, or Infinity fall back to
+  //     the built-in defaults.
+  const envTimeout = Number(process.env.HF_DOWNLOAD_TIMEOUT_MS);
+  const envMaxAttempts = Number(process.env.HF_MAX_ATTEMPTS);
+  const resolvedTimeout =
+    Number.isFinite(envTimeout) && envTimeout > 0
+      ? Math.min(envTimeout, HF_MAX_TIMEOUT_MS)
+      : HF_DOWNLOAD_TIMEOUT_MS;
+  const resolvedMaxAttempts =
+    Number.isFinite(envMaxAttempts) && envMaxAttempts > 0
+      ? Math.min(Math.floor(envMaxAttempts), HF_MAX_ATTEMPTS_CAP)
+      : HF_MAX_ATTEMPTS;
+  const {
+    maxAttempts = resolvedMaxAttempts,
+    baseDelayMs = HF_BASE_DELAY_MS,
+    timeoutMs = resolvedTimeout,
+    circuit = hfDownloadCircuit,
+    onRetry,
+  } = options;
+  if (circuit.isOpen()) {
+    const secsUntilReset = Math.ceil(
+      (circuit.resetTimeoutMs - (Date.now() - circuit.lastFailureAt)) / 1000,
+    );
+    throw new Error(
+      `${CIRCUIT_OPEN_TAG}: HuggingFace download circuit is open after repeated network failures` +
+        (secsUntilReset > 0 ? ` — will reset in ~${secsUntilReset}s` : ''),
+    );
+  }
+
+  let lastError: Error = new Error('unknown error');
+
+  for (let attempt = 0; attempt < maxAttempts; attempt++) {
+    try {
+      const result = await withDownloadTimeout(fn, timeoutMs);
+      circuit.recordSuccess();
+      return result;
+    } catch (err) {
+      lastError = err instanceof Error ? err : new Error(String(err));
+
+      if (!isNetworkFetchError(lastError.message)) {
+        // Non-network error (e.g. CUDA unavailable) — propagate without retry
+        throw lastError;
+      }
+
+      circuit.recordFailure();
+
+      if (circuit.isOpen()) {
+        // Circuit just tripped — fail fast, no more retries
+        throw new Error(
+          `${CIRCUIT_OPEN_TAG}: HuggingFace download circuit opened after ${circuit.failureThreshold} consecutive failures`,
+        );
+      }
+
+      if (attempt < maxAttempts - 1) {
+        const delay = baseDelayMs * Math.pow(2, attempt);
+        onRetry?.(attempt + 1, maxAttempts, lastError);
+        await sleep(delay);
+      }
+    }
+  }
+
+  // All retries exhausted — throw the last network error so isNetworkFetchError
+  // patterns in the calling code still match and surface HF_ENDPOINT guidance.
+  throw lastError;
+}
diff --git a/gitnexus/src/core/git-staleness.ts b/gitnexus/src/core/git-staleness.ts
index 96f70ddd6..c90cef85e 100644
--- a/gitnexus/src/core/git-staleness.ts
+++ b/gitnexus/src/core/git-staleness.ts
@@ -3,11 +3,14 @@
  * Lives in core/ so application code does not depend on the MCP package layer.
  */
 
-import { execFileSync } from 'node:child_process';
+import { execFile, execFileSync } from 'node:child_process';
+import { promisify } from 'node:util';
 import path from 'path';
 import { readRegistry, type RegistryEntry, type CwdMatch } from '../storage/repo-manager.js';
 import { findGitRootByDotGit, getCurrentCommit, getRemoteUrl } from '../storage/git.js';
 
+const execFileAsync = promisify(execFile);
+
 export interface StalenessInfo {
   isStale: boolean;
   commitsBehind: number;
@@ -41,6 +44,39 @@ export function checkStaleness(repoPath: string, lastCommit: string): StalenessI
   }
 }
 
+/**
+ * Async variant of {@link checkStaleness} — spawns git as a child process
+ * instead of blocking the event loop.  Used by `listRepos()` to check many
+ * repos in parallel (issue #1363: 200 repos × sync spawn ≈ 50 s).
+ */
+export async function checkStalenessAsync(
+  repoPath: string,
+  lastCommit: string,
+): Promise<StalenessInfo> {
+  try {
+    // Note: promisified execFile captures stdout/stderr by default (no stdio option needed,
+    // unlike the sync variant which requires explicit stdio: ['pipe','pipe','pipe']).
+    const { stdout } = await execFileAsync('git', ['rev-list', '--count', `${lastCommit}..HEAD`], {
+      cwd: repoPath,
+      encoding: 'utf-8',
+    });
+
+    const commitsBehind = parseInt(stdout.trim(), 10) || 0;
+
+    if (commitsBehind > 0) {
+      return {
+        isStale: true,
+        commitsBehind,
+        hint: `⚠️ Index is ${commitsBehind} commit${commitsBehind > 1 ? 's' : ''} behind HEAD. Run analyze tool to update.`,
+      };
+    }
+
+    return { isStale: false, commitsBehind: 0 };
+  } catch {
+    return { isStale: false, commitsBehind: 0 };
+  }
+}
+
 /**
  * Compare a sibling-clone HEAD against an indexed `lastCommit`. Returns
  * `undefined` when the indexed commit is not reachable from the sibling
diff --git a/gitnexus/src/core/group/config-parser.ts b/gitnexus/src/core/group/config-parser.ts
index 73a9021b9..29c868171 100644
--- a/gitnexus/src/core/group/config-parser.ts
+++ b/gitnexus/src/core/group/config-parser.ts
@@ -4,9 +4,26 @@ import type { GroupConfig, GroupManifestLink, ContractType, ContractRole } from
 const _require = createRequire(import.meta.url);
 const yaml = _require('js-yaml') as typeof import('js-yaml');
 
-const VALID_CONTRACT_TYPES: ContractType[] = ['http', 'grpc', 'thrift', 'topic', 'lib', 'custom'];
+const VALID_CONTRACT_TYPES: ContractType[] = [
+  'http',
+  'grpc',
+  'thrift',
+  'topic',
+  'lib',
+  'custom',
+  'include',
+];
 const VALID_ROLES: ContractRole[] = ['provider', 'consumer'];
 
+// Defaults matter for backward compatibility: any group.yaml that omits a
+// `detect.<type>` key inherits its value from this constant. Adding a new
+// extractor that defaults to `true` silently changes the behavior of every
+// existing group on the next sync. New extractors must default to `false`
+// (opt-in) so operators consciously enable them via group.yaml.
+//
+// `includes`: opt-in. The C/C++ IncludeExtractor (PR #1156) ships disabled by
+// default; enable with `detect.includes: true` for groups containing C/C++
+// repos that need cross-repo header tracking.
 const DEFAULT_DETECT = {
   http: true,
   grpc: true,
@@ -14,6 +31,7 @@ const DEFAULT_DETECT = {
   topics: true,
   shared_libs: true,
   embedding_fallback: true,
+  includes: false,
   workspace_deps: false,
 };
 
diff --git a/gitnexus/src/core/group/cross-impact.ts b/gitnexus/src/core/group/cross-impact.ts
index f8625cdc5..eab942a62 100644
--- a/gitnexus/src/core/group/cross-impact.ts
+++ b/gitnexus/src/core/group/cross-impact.ts
@@ -91,6 +91,25 @@ function clampCrossDepth(raw: unknown): { depth: number; warning?: string } {
   return { depth: d };
 }
 
+/**
+ * Clamp the impact timeout to a sane bounded range. Callers can feed this
+ * via tool params, so an unclamped value lets a single request hold a
+ * timer slot for an arbitrarily long duration (CodeQL js/resource-
+ * exhaustion). 100ms lower bound preserves test-suite scenarios that
+ * exercise tight timeouts; 5min upper bound is well above any legitimate
+ * single-impact compute. Applied at the validate boundary so the
+ * downstream `deadline` (Date.now() + timeoutMs) and the local-leg
+ * `setTimeout` see the same clamped value — earlier shapes had a 1hr
+ * outer cap and a 5min inner clamp that disagreed.
+ */
+export const IMPACT_TIMEOUT_MIN_MS = 100;
+export const IMPACT_TIMEOUT_MAX_MS = 5 * 60 * 1_000;
+
+export function clampTimeout(timeoutMs: number): number {
+  if (!Number.isFinite(timeoutMs) || timeoutMs <= 0) return IMPACT_TIMEOUT_MIN_MS;
+  return Math.min(IMPACT_TIMEOUT_MAX_MS, Math.max(IMPACT_TIMEOUT_MIN_MS, Math.trunc(timeoutMs)));
+}
+
 export function validateGroupImpactParams(params: Record<string, unknown>):
   | {
       ok: true;
@@ -143,13 +162,19 @@ export function validateGroupImpactParams(params: Record<string, unknown>):
   const service = normalizeServicePrefix(params.service);
   const subgroup = typeof params.subgroup === 'string' ? params.subgroup : undefined;
 
-  let timeoutMs =
+  // Clamp at the validate boundary so the downstream `deadline` (line
+  // ~366) and `safeLocalImpact`'s `setTimeout` both see a single
+  // bounded value. Without this, the outer deadline budgeted Phase-2
+  // cross-repo fanout up to 1hr while only the inner setTimeout was
+  // capped to 5min — the two halves of CodeQL #184's mitigation
+  // disagreed.
+  const rawTimeoutMs =
     typeof params.timeoutMs === 'number' && params.timeoutMs > 0
       ? params.timeoutMs
       : typeof params.timeout === 'number' && params.timeout > 0
         ? params.timeout
         : DEFAULT_LOCAL_IMPACT_TIMEOUT_MS;
-  if (timeoutMs > 3_600_000) timeoutMs = 3_600_000;
+  const timeoutMs = clampTimeout(rawTimeoutMs);
 
   return {
     ok: true,
@@ -191,12 +216,13 @@ async function safeLocalImpact(
   impactParams: Parameters<GroupToolPort['impact']>[1],
   timeoutMs: number,
 ): Promise<{ value: unknown; timedOut: boolean }> {
+  const safeTimeoutMs = clampTimeout(timeoutMs);
   let timer: ReturnType<typeof setTimeout> | undefined;
   const impactP = port.impact(repo, impactParams).catch((err) => ({
     error: err instanceof Error ? err.message : String(err),
   }));
   const timeoutP = new Promise<'timeout'>((resolve) => {
-    timer = setTimeout(() => resolve('timeout'), timeoutMs);
+    timer = setTimeout(() => resolve('timeout'), safeTimeoutMs);
   });
   const won = await Promise.race([
     impactP.then((v) => ({ tag: 'impact' as const, v })),
@@ -212,6 +238,65 @@ async function safeLocalImpact(
   return { value: won.v, timedOut: false };
 }
 
+/**
+ * Race a single Phase-2 `impactByUid` call against a remaining-budget
+ * timer. The Codex adversarial review on PR #1331 surfaced that the
+ * fanout loop only checked `Date.now() > deadline` *between* neighbor
+ * calls — once `await port.impactByUid(...)` was reached, a hung
+ * neighbor could pin the request indefinitely, and slow neighbors
+ * could compound past the 5-min `IMPACT_TIMEOUT_MAX_MS` cap.
+ *
+ * This helper wraps each call: a `setTimeout(remainingMs)` aborts an
+ * `AbortController` whose signal is forwarded to `impactByUid`, and a
+ * `Promise.race` resolves to `{ timedOut: true }` when the timer
+ * fires before the call completes. Implementors that ignore the
+ * signal (current local backend) still see their await resolved by
+ * the race; full cooperative cancellation inside the BFS is a future
+ * follow-up. On rejection, the value is `null` (matching the
+ * fanout's existing `if (fan == null)` truncation contract).
+ *
+ * Exported for direct unit testing — the helper IS the load-bearing
+ * mitigation surface, so the U3 regression test pins it directly
+ * rather than driving the full `runGroupImpact` path.
+ */
+export async function safeNeighborImpact(
+  port: GroupToolPort,
+  repoId: string,
+  uid: string,
+  direction: string,
+  opts: {
+    maxDepth: number;
+    relationTypes: string[];
+    minConfidence: number;
+    includeTests: boolean;
+  },
+  remainingMs: number,
+): Promise<{ value: unknown; timedOut: boolean }> {
+  const controller = new AbortController();
+  let timer: ReturnType<typeof setTimeout> | undefined;
+  const callP = port
+    .impactByUid(repoId, uid, direction, { ...opts, signal: controller.signal })
+    .catch(() => null);
+  const timeoutP = new Promise<'timeout'>((resolve) => {
+    timer = setTimeout(
+      () => {
+        controller.abort();
+        resolve('timeout');
+      },
+      Math.max(0, remainingMs),
+    );
+  });
+  const won = await Promise.race([
+    callP.then((v) => ({ tag: 'impact' as const, v })),
+    timeoutP.then(() => ({ tag: 'timeout' as const })),
+  ]);
+  if (timer !== undefined) clearTimeout(timer);
+  if (won.tag === 'timeout') {
+    return { value: null, timedOut: true };
+  }
+  return { value: won.v, timedOut: false };
+}
+
 export function collectImpactSymbolUids(
   local: unknown,
   servicePrefix: string | undefined,
@@ -476,7 +561,8 @@ export async function runGroupImpact(
       if (seen.has(key)) continue;
       seen.add(key);
 
-      if (Date.now() > deadline) {
+      const remainingMs = deadline - Date.now();
+      if (remainingMs <= 0) {
         truncatedRepos.push(n.neighborRepo);
         continue;
       }
@@ -492,13 +578,25 @@ export async function runGroupImpact(
         continue;
       }
 
-      const fan = await deps.port.impactByUid(neighborHandle.id, n.neighborUid, direction, {
-        maxDepth,
-        relationTypes: relationTypes ?? [],
-        minConfidence,
-        includeTests,
-      });
-      if (fan == null) {
+      // Phase-2 hardening: race each impactByUid against a per-call
+      // timeout derived from the remaining budget. Without this wrap a
+      // single hung neighbor would pin the request past the clamped
+      // timeout, which Codex's adversarial review on PR #1331 flagged
+      // as the still-open half of CodeQL #184 / js/resource-exhaustion.
+      const { value: fan, timedOut: neighborTimedOut } = await safeNeighborImpact(
+        deps.port,
+        neighborHandle.id,
+        n.neighborUid,
+        direction,
+        {
+          maxDepth,
+          relationTypes: relationTypes ?? [],
+          minConfidence,
+          includeTests,
+        },
+        remainingMs,
+      );
+      if (neighborTimedOut || fan == null) {
         truncatedRepos.push(n.neighborRepo);
         continue;
       }
diff --git a/gitnexus/src/core/group/extractors/include-extractor.ts b/gitnexus/src/core/group/extractors/include-extractor.ts
new file mode 100644
index 000000000..7bbfd61ed
--- /dev/null
+++ b/gitnexus/src/core/group/extractors/include-extractor.ts
@@ -0,0 +1,610 @@
+import * as path from 'node:path';
+import * as fs from 'node:fs/promises';
+import { glob } from 'glob';
+import Parser from 'tree-sitter';
+import C from 'tree-sitter-c';
+import Cpp from 'tree-sitter-cpp';
+import type { ContractExtractor, CypherExecutor } from '../contract-extractor.js';
+import type { ExtractedContract, RepoHandle } from '../types.js';
+import { readSafe } from './fs-utils.js';
+import { buildSuffixIndex, type SuffixIndex } from '../../ingestion/import-resolvers/utils.js';
+import { createIgnoreFilter } from '../../../config/ignore-service.js';
+import { getMaxFileSizeBytes } from '../../ingestion/utils/max-file-size.js';
+import { logger } from '../../logger.js';
+
+/**
+ * Cross-repo C/C++ `#include` dependency extractor.
+ *
+ * **Provider side:** registers every `.h/.hpp/.hxx/.hh` file in the repo
+ * as a provider contract with `include::<relative-path>`.
+ *
+ * **Consumer side:** parses all C/C++ source/header files for `#include "…"`
+ * directives, attempts suffix-based resolution against the repo's own file
+ * list (reusing the same algorithm as the single-repo ingestion pipeline),
+ * and emits unresolved include paths as consumer contracts.
+ *
+ * Matching: a consumer's `include::map/base/dice_map_view.h` in repo A
+ * matches a provider's `include::map/base/dice_map_view.h` in repo B via
+ * exact contract-id equality in `runExactMatch`.
+ */
+
+// ---------- constants ----------
+
+const HEADER_EXTENSIONS = new Set(['.h', '.hpp', '.hxx', '.hh']);
+
+// Source = headers (provider-eligible) ∪ implementation files (.c/.cpp/.cc/.cxx).
+// Spread keeps the subset relationship explicit so a future contributor adding
+// a new header extension to HEADER_EXTENSIONS does not have to remember to
+// also add it here.
+const SOURCE_EXTENSIONS = new Set<string>([...HEADER_EXTENSIONS, '.c', '.cpp', '.cc', '.cxx']);
+
+const INCLUDE_QUERY_SRC = '(preproc_include path: (_) @import.source) @import';
+
+/**
+ * Well-known C/C++ standard library headers that can appear in `#include "…"`
+ * form (some projects use quotes for system headers).
+ */
+const SYSTEM_HEADERS = new Set([
+  // C standard
+  'assert.h',
+  'complex.h',
+  'ctype.h',
+  'errno.h',
+  'fenv.h',
+  'float.h',
+  'inttypes.h',
+  'iso646.h',
+  'limits.h',
+  'locale.h',
+  'math.h',
+  'setjmp.h',
+  'signal.h',
+  'stdalign.h',
+  'stdarg.h',
+  'stdatomic.h',
+  'stdbool.h',
+  'stddef.h',
+  'stdint.h',
+  'stdio.h',
+  'stdlib.h',
+  'stdnoreturn.h',
+  'string.h',
+  'tgmath.h',
+  'threads.h',
+  'time.h',
+  'uchar.h',
+  'wchar.h',
+  'wctype.h',
+  // C++ standard (extensionless)
+  'algorithm',
+  'any',
+  'array',
+  'atomic',
+  'barrier',
+  'bit',
+  'bitset',
+  'cassert',
+  'cctype',
+  'cerrno',
+  'cfenv',
+  'cfloat',
+  'charconv',
+  'chrono',
+  'cinttypes',
+  'climits',
+  'clocale',
+  'cmath',
+  'codecvt',
+  'compare',
+  'complex',
+  'concepts',
+  'condition_variable',
+  'coroutine',
+  'csetjmp',
+  'csignal',
+  'cstdarg',
+  'cstddef',
+  'cstdint',
+  'cstdio',
+  'cstdlib',
+  'cstring',
+  'ctime',
+  'cuchar',
+  'cwchar',
+  'cwctype',
+  'deque',
+  'exception',
+  'execution',
+  'expected',
+  'filesystem',
+  'format',
+  'forward_list',
+  'fstream',
+  'functional',
+  'future',
+  'generator',
+  'initializer_list',
+  'iomanip',
+  'ios',
+  'iosfwd',
+  'iostream',
+  'istream',
+  'iterator',
+  'latch',
+  'limits',
+  'list',
+  'locale',
+  'map',
+  'mdspan',
+  'memory',
+  'memory_resource',
+  'mutex',
+  'new',
+  'numbers',
+  'numeric',
+  'optional',
+  'ostream',
+  'print',
+  'queue',
+  'random',
+  'ranges',
+  'ratio',
+  'regex',
+  'scoped_allocator',
+  'semaphore',
+  'set',
+  'shared_mutex',
+  'source_location',
+  'span',
+  'spanstream',
+  'sstream',
+  'stack',
+  'stacktrace',
+  'stdexcept',
+  'stdfloat',
+  'stop_token',
+  'streambuf',
+  'string',
+  'string_view',
+  'strstream',
+  'syncstream',
+  'system_error',
+  'thread',
+  'tuple',
+  'type_traits',
+  'typeindex',
+  'typeinfo',
+  'unordered_map',
+  'unordered_set',
+  'utility',
+  'valarray',
+  'variant',
+  'vector',
+  'version',
+]);
+
+/** Path prefixes that indicate system/kernel headers. */
+const SYSTEM_PATH_PREFIXES = [
+  'sys/',
+  'net/',
+  'netinet/',
+  'arpa/',
+  'linux/',
+  'asm/',
+  'bits/',
+  'gnu/',
+  'mach/',
+  'machine/',
+  'xlocale/',
+];
+
+/** Regex fallback for files that exceed tree-sitter's 32 KB parse limit. */
+const INCLUDE_REGEX = /^[ \t]*#\s*include\s*"([^"]+)"/gm;
+
+// ---------- helpers ----------
+
+/**
+ * Normalize an include path to a canonical lowercase forward-slash form.
+ *
+ * IMPORTANT — case-folding caveat (PR #1156 review finding #3):
+ *   Header paths are lowercased so consumer `#include "Foo/Bar.h"` and
+ *   provider file `Foo/Bar.h` normalize to the same contract-id. This is
+ *   the right trade-off on case-insensitive filesystems (macOS, Windows)
+ *   but on case-sensitive Linux filesystems two distinct headers `Foo.h`
+ *   and `foo.h` in the same repo will collide onto the same provider
+ *   contract-id; only one survives `dedupe()`. The gain (reliable
+ *   cross-platform matching) outweighs the cost (extremely rare header
+ *   casing collisions inside a single repo).
+ */
+function normalizeIncludePath(raw: string): string {
+  return raw.replace(/\\/g, '/').replace(/^\.\//, '').replace(/\/+/g, '/').toLowerCase();
+}
+
+/**
+ * Strip C/C++ block comments from a source blob. Used only by the
+ * regex-fallback path to avoid emitting consumer contracts for
+ * commented-out #include directives. Line comments (`// …`) cannot hide
+ * #include directives because the regex anchors on start-of-line.
+ * See PR #1156 review finding #5.
+ */
+function stripBlockComments(src: string): string {
+  return src.replace(/\/\*[\s\S]*?\*\//g, '');
+}
+
+function isAngleBracketInclude(rawNodeText: string): boolean {
+  const trimmed = rawNodeText.trim();
+  return trimmed.startsWith('<') && trimmed.endsWith('>');
+}
+
+function isSystemHeader(cleanedPath: string): boolean {
+  // Check well-known standard headers
+  if (SYSTEM_HEADERS.has(cleanedPath)) return true;
+  // Check system path prefixes
+  const lower = cleanedPath.toLowerCase();
+  return SYSTEM_PATH_PREFIXES.some((prefix) => lower.startsWith(prefix));
+}
+
+function isHeaderFile(filePath: string): boolean {
+  return HEADER_EXTENSIONS.has(path.extname(filePath).toLowerCase());
+}
+
+function getLanguageForFile(filePath: string): unknown | null {
+  const ext = path.extname(filePath).toLowerCase();
+  switch (ext) {
+    case '.c':
+    case '.h':
+      return C;
+    case '.cpp':
+    case '.cc':
+    case '.cxx':
+    case '.hpp':
+    case '.hxx':
+    case '.hh':
+      return Cpp;
+    default:
+      return null;
+  }
+}
+
+/**
+ * Check whether an include path resolves to a file inside the local repo.
+ *
+ * Uses *exact full-path* matching on the suffix index — we never accept a
+ * truncated suffix match. For `#include "foo/bar.h"` this checks:
+ *   (a) a file whose path ends with the full `foo/bar.h`
+ *   (b) if the include omitted the extension, a file whose path ends with
+ *       the include + one of the C/C++ header extensions
+ *
+ * Returns `true` when a local file matches — caller should suppress the
+ * cross-repo consumer contract.
+ *
+ * See PR #1156 review finding #4 (suffixResolve ambiguity).
+ */
+function isLocalInclude(cleaned: string, suffixIndex: SuffixIndex): boolean {
+  const candidates = [cleaned];
+  if (!/\.[a-zA-Z0-9]+$/.test(cleaned)) {
+    for (const ext of ['.h', '.hpp', '.hxx', '.hh']) candidates.push(cleaned + ext);
+  }
+  for (const c of candidates) {
+    if (suffixIndex.get(c) || suffixIndex.getInsensitive(c)) return true;
+  }
+  return false;
+}
+
+// ---------- main class ----------
+
+export class IncludeExtractor implements ContractExtractor {
+  type = 'include' as const;
+
+  /**
+   * Always returns `true`. NOT called by `sync.ts`, which gates extraction via
+   * `config.detect.includes` instead (see `sync.ts:174`). Kept solely to satisfy
+   * the `ContractExtractor` interface so the type stays uniform across extractors.
+   */
+  async canExtract(_repo: RepoHandle): Promise<boolean> {
+    return true;
+  }
+
+  async extract(
+    dbExecutor: CypherExecutor | null,
+    repoPath: string,
+    _repo: RepoHandle,
+  ): Promise<ExtractedContract[]> {
+    // 1. Build the local file list using the same discovery as ingestion
+    //    (createIgnoreFilter + getMaxFileSizeBytes). This guarantees the
+    //    universe of provider/consumer paths matches the universe of File
+    //    nodes in the LadybugDB graph — so no cross-link points at a UID
+    //    that group impact cannot fan out to.
+    //    (PR #1156 Codex follow-up: discovery aligned with ingestion.)
+    const allFiles = await this.discoverIndexableFiles(repoPath);
+    const normalizedFiles = allFiles.map((f) => f.replace(/\\/g, '/'));
+    const suffixIndex = buildSuffixIndex(normalizedFiles, allFiles);
+
+    // 2. Provider: register all header files
+    const providers = await this.extractProviders(dbExecutor, repoPath, allFiles);
+
+    // 3. Consumer: filter the shared discovery list for source extensions
+    //    and parse #include directives in those files.
+    const sourceFiles = allFiles.filter((f) =>
+      SOURCE_EXTENSIONS.has(path.extname(f).toLowerCase()),
+    );
+    const consumers = await this.extractConsumers(repoPath, sourceFiles, suffixIndex);
+
+    return this.dedupe([...providers, ...consumers]);
+  }
+
+  /**
+   * Discover repo-relative file paths using exactly the same rules the
+   * ingestion pipeline uses (`walkRepositoryPaths` in
+   * `gitnexus/src/core/ingestion/filesystem-walker.ts`):
+   *   - `createIgnoreFilter` honors `.gitignore`, `.gitnexusignore`, the
+   *     hardcoded ignore list, and `.gitnexusignore` last-match-wins
+   *     negation.
+   *   - `getMaxFileSizeBytes()` drops files larger than the cap so we
+   *     never emit `File:<rel>` UIDs for files ingestion would skip.
+   *
+   * Uses sequential stat — there is no `READ_CONCURRENCY` batching here
+   * because group sync runs at startup-time, not the ingestion hot path,
+   * and parallelism gains are not worth the import-graph weight.
+   *
+   * MAINTENANCE: if `walkRepositoryPaths` changes its glob options, ignore
+   * filter shape, or size-cap logic, mirror those changes here. The two
+   * implementations exist because the consumers need different return
+   * shapes (string[] vs ScannedFile[]) and different concurrency, but
+   * they MUST agree on which files are reachable — that is what makes
+   * `File:<rel>` UIDs in cross-links correspond to graph File nodes.
+   */
+  private async discoverIndexableFiles(repoPath: string): Promise<string[]> {
+    const ignoreFilter = await createIgnoreFilter(repoPath);
+    const maxFileSizeBytes = getMaxFileSizeBytes();
+
+    const candidates = await glob('**/*', {
+      cwd: repoPath,
+      nodir: true,
+      dot: false,
+      ignore: ignoreFilter,
+    });
+
+    const survivors: string[] = [];
+    for (const rel of candidates) {
+      try {
+        const stat = await fs.stat(path.join(repoPath, rel));
+        if (stat.size > maxFileSizeBytes) continue;
+        survivors.push(rel);
+      } catch (err) {
+        // ENOENT is the documented benign race (glob enumerated a file
+        // that was deleted before we stat'd it — same race
+        // walkRepositoryPaths absorbs via Promise.allSettled). Anything
+        // else (EACCES, EMFILE, EIO) deserves a warning so an operator
+        // can spot a permission/resource problem instead of silently
+        // shipping fewer contracts than expected.
+        const code = (err as NodeJS.ErrnoException | undefined)?.code;
+        if (code !== 'ENOENT') {
+          logger.warn(
+            { err: (err as Error).message, file: rel, repoPath },
+            '⚠️ IncludeExtractor: stat failed during discovery; skipping file',
+          );
+        }
+      }
+    }
+    return survivors;
+  }
+
+  // ---------- provider extraction ----------
+
+  private async extractProviders(
+    dbExecutor: CypherExecutor | null,
+    repoPath: string,
+    allFiles: string[],
+  ): Promise<ExtractedContract[]> {
+    // Strategy A: graph-assisted
+    if (dbExecutor) {
+      const graphProviders = await this.extractProvidersGraph(dbExecutor, repoPath);
+      if (graphProviders.length > 0) return graphProviders;
+    }
+    // Strategy B: filesystem fallback
+    return this.extractProvidersFallback(repoPath, allFiles);
+  }
+
+  private async extractProvidersGraph(
+    db: CypherExecutor,
+    repoPath: string,
+  ): Promise<ExtractedContract[]> {
+    try {
+      const rows = await db(
+        `MATCH (f:File)
+         WHERE f.filePath =~ '.*\\\\.(h|hpp|hxx|hh)$'
+         RETURN f.filePath AS filePath, f.id AS fileId`,
+      );
+      // gitnexus analyze stores absolute paths in the File.filePath column.
+      // Provider contract IDs MUST be repo-relative — otherwise the consumer
+      // emits `include::map/base/view.h` and the provider emits
+      // `include::/abs/path/to/repo/map/base/view.h`, which never match
+      // through runExactMatch and the cross-link silently disappears.
+      // (PR #1156 follow-up review: graph provider absolute-path bug.)
+      const normalizedRepoPath = path.resolve(repoPath);
+      const out: ExtractedContract[] = [];
+      for (const r of rows) {
+        if (typeof r.filePath !== 'string' || !r.filePath) continue;
+        const absolute = r.filePath as string;
+        const rel = path.relative(normalizedRepoPath, absolute);
+        // Skip rows that resolve outside the repo (e.g., system headers
+        // somehow indexed, or stale absolute paths from a different machine).
+        // path.relative returns a `..`-prefixed path or an absolute path
+        // when the target is outside the base — both are wrong for our IDs.
+        if (!rel || rel.startsWith('..') || path.isAbsolute(rel)) continue;
+        const normalizedRel = rel.replace(/\\/g, '/');
+        out.push({
+          contractId: `include::${normalizeIncludePath(normalizedRel)}`,
+          type: 'include' as const,
+          role: 'provider' as const,
+          symbolUid: String(r.fileId ?? ''),
+          symbolRef: { filePath: normalizedRel, name: path.basename(normalizedRel) },
+          symbolName: path.basename(normalizedRel),
+          confidence: 1.0,
+          meta: { source: 'graph' },
+        });
+      }
+      return out;
+    } catch {
+      return [];
+    }
+  }
+
+  private extractProvidersFallback(_repoPath: string, allFiles: string[]): ExtractedContract[] {
+    return allFiles
+      .filter((f) => isHeaderFile(f))
+      .map((f) => {
+        const filePath = f.replace(/\\/g, '/');
+        return {
+          contractId: `include::${normalizeIncludePath(filePath)}`,
+          type: 'include' as const,
+          role: 'provider' as const,
+          symbolUid: `File:${filePath}`,
+          symbolRef: { filePath, name: path.basename(filePath) },
+          symbolName: path.basename(filePath),
+          confidence: 0.95,
+          meta: { source: 'filesystem' },
+        };
+      });
+  }
+
+  // ---------- consumer extraction ----------
+
+  private async extractConsumers(
+    repoPath: string,
+    sourceFiles: string[],
+    suffixIndex: SuffixIndex,
+  ): Promise<ExtractedContract[]> {
+    const parser = new Parser();
+    const out: ExtractedContract[] = [];
+    // Compile the include query once per grammar to avoid re-compilation per file
+    const queryCache = new Map<unknown, Parser.Query>();
+
+    for (const rel of sourceFiles) {
+      const lang = getLanguageForFile(rel);
+      if (!lang) continue;
+
+      const content = readSafe(repoPath, rel);
+      if (!content) continue;
+
+      let query = queryCache.get(lang);
+      if (!query) {
+        try {
+          query = new Parser.Query(lang, INCLUDE_QUERY_SRC);
+          queryCache.set(lang, query);
+        } catch {
+          continue;
+        }
+      }
+
+      // Collect raw include paths: tree-sitter first, regex fallback for large files.
+      // `extractionSource` is stamped on each emitted consumer contract so
+      // regex-fallback contracts stay auditable post-hoc (PR #1156 review finding #6).
+      let rawIncludes: string[];
+      let extractionSource: 'tree_sitter' | 'regex_fallback';
+      try {
+        parser.setLanguage(lang);
+        const tree = parser.parse(content);
+        let matches: Parser.QueryMatch[];
+        try {
+          matches = query.matches(tree.rootNode);
+        } catch {
+          matches = [];
+        }
+        rawIncludes = [];
+        extractionSource = 'tree_sitter';
+        for (const match of matches) {
+          const sourceNode = match.captures.find((c) => c.name === 'import.source');
+          if (!sourceNode) continue;
+          const rawText = sourceNode.node.text;
+          if (isAngleBracketInclude(rawText)) continue;
+          const cleaned = rawText.replace(/['"<>]/g, '');
+          if (cleaned && cleaned.length <= 2048) rawIncludes.push(cleaned);
+        }
+      } catch {
+        // tree-sitter failed (e.g. file > 32 KB) — fall back to regex.
+        // Strip block comments first so we don't emit a consumer contract
+        // for a commented-out #include (PR #1156 review finding #5).
+        rawIncludes = [];
+        extractionSource = 'regex_fallback';
+        const scanTarget = stripBlockComments(content);
+        INCLUDE_REGEX.lastIndex = 0;
+        let m: RegExpExecArray | null;
+        while ((m = INCLUDE_REGEX.exec(scanTarget)) !== null) {
+          if (m[1] && m[1].length <= 2048) rawIncludes.push(m[1]);
+        }
+      }
+
+      for (const cleaned of rawIncludes) {
+        // Filter: skip known system headers and system path prefixes
+        if (isSystemHeader(cleaned)) continue;
+
+        // Skip relative-up includes: `#include "../include/foo.h"` is
+        // almost always an intra-repo reference. The suffix index is built
+        // from repo-relative paths, so isLocalInclude can never match
+        // `../foo.h`, and emitting it as a consumer contract just pollutes
+        // the registry with an entry no provider can ever satisfy.
+        // (PR #1156 follow-up review: `../` relative includes produce
+        // spurious consumer contracts.)
+        if (cleaned.startsWith('../') || cleaned.startsWith('..\\')) continue;
+
+        // Skip macro-style includes: `#include PLATFORM_HEADER` parses as an
+        // identifier under tree-sitter's `(_) @import.source` wildcard. The
+        // identifier text passes the strip/clean step unchanged, so without
+        // this guard we would emit `include::platform_header` as a consumer
+        // contract — and no provider in any repo will ever expose a contract
+        // for a macro identifier (no file is named `PLATFORM_HEADER`). The
+        // contract would sit permanently orphaned in the registry. Real
+        // header references always contain a path separator (`/`, `\`) or an
+        // extension dot (`foo.h`), so an absent both is a reliable signal we
+        // are looking at a macro identifier. (PR #1156 follow-up review:
+        // macro includes emit orphaned consumer contracts.)
+        if (!/[./\\]/.test(cleaned)) continue;
+
+        // Local resolution (PR #1156 review finding #4): only accept an
+        // exact-suffix match on the *full* include path. The generic
+        // suffixResolve() iterates all truncated suffixes, which would
+        // silently suppress a cross-repo `#include "map/base/view.h"`
+        // when the local repo has any `internal/view.h` — a realistic
+        // false-negative in large C++ codebases. Here we only resolve
+        // locally if a file path ends with the complete include string
+        // (optionally re-appending one of the C/C++ header extensions
+        // when the include already omits it).
+        if (isLocalInclude(cleaned, suffixIndex)) continue;
+
+        // Unresolved: emit as consumer contract
+        const normalizedRel = rel.replace(/\\/g, '/');
+        out.push({
+          contractId: `include::${normalizeIncludePath(cleaned)}`,
+          type: 'include' as const,
+          role: 'consumer' as const,
+          symbolUid: `File:${normalizedRel}`,
+          symbolRef: { filePath: normalizedRel, name: cleaned },
+          symbolName: cleaned,
+          confidence: 0.85,
+          meta: {
+            source: extractionSource,
+            includePath: cleaned,
+          },
+        });
+      }
+    }
+
+    return out;
+  }
+
+  // ---------- deduplication ----------
+
+  private dedupe(items: ExtractedContract[]): ExtractedContract[] {
+    const seen = new Set<string>();
+    const out: ExtractedContract[] = [];
+    for (const c of items) {
+      const k = `${c.contractId}|${c.role}|${c.symbolRef.filePath}`;
+      if (seen.has(k)) continue;
+      seen.add(k);
+      out.push(c);
+    }
+    return out;
+  }
+}
diff --git a/gitnexus/src/core/group/extractors/manifest-extractor.ts b/gitnexus/src/core/group/extractors/manifest-extractor.ts
index 2af3db595..f4d0f77cf 100644
--- a/gitnexus/src/core/group/extractors/manifest-extractor.ts
+++ b/gitnexus/src/core/group/extractors/manifest-extractor.ts
@@ -274,6 +274,14 @@ export class ManifestExtractor {
            LIMIT 1`,
           { contract: link.contract },
         );
+      } else if (link.type === 'include') {
+        rows = await executor(
+          `MATCH (f:File) WHERE f.filePath = $contract
+           RETURN f.id AS uid, f.name AS name, f.filePath AS filePath
+           ORDER BY f.filePath ASC
+           LIMIT 1`,
+          { contract: link.contract },
+        );
       } else if (link.type === 'custom') {
         // Workspace extractors produce qualified contracts like "mathlex::Expression".
         // Graph nodes store the unqualified symbol name ("Expression"), so strip
@@ -358,6 +366,8 @@ export class ManifestExtractor {
         return `lib::${contract}`;
       case 'custom':
         return `custom::${contract}`;
+      case 'include':
+        return `include::${contract}`;
       default: {
         const _exhaustive: never = type;
         throw new Error(`Unhandled ContractType: ${String(_exhaustive)}`);
diff --git a/gitnexus/src/core/group/extractors/rust-workspace-extractor.ts b/gitnexus/src/core/group/extractors/rust-workspace-extractor.ts
index 63fe7ea82..d58c3e08f 100644
--- a/gitnexus/src/core/group/extractors/rust-workspace-extractor.ts
+++ b/gitnexus/src/core/group/extractors/rust-workspace-extractor.ts
@@ -31,6 +31,32 @@ interface ImportedSymbol {
   filePath: string;
 }
 
+/**
+ * Linear-time `[package].name = "..."` lookup. The previous regex
+ * `^\[package\]\s*\n(?:[^\[]*?\n)*?name\s*=\s*"([^"]+)"` had a nested
+ * lazy quantifier on `\n` that CodeQL js/redos flagged as exponential
+ * on inputs like `[package]\n` + many bare `\n`. We walk lines
+ * explicitly: scan from the first `[package]` header until we hit the
+ * next `[...]` section header, looking for the `name = "..."` line.
+ * O(n) with the line count.
+ *
+ * Exported so the U8 ReDoS regression test can drive the production
+ * line-walk directly with adversarial fixtures (multi-line strings,
+ * trailing sections, etc.) instead of duplicating it inline.
+ */
+export function parseCargoPackageName(content: string): string | null {
+  const lines = content.split('\n');
+  const packageStart = lines.findIndex((l) => l.trim() === '[package]');
+  if (packageStart < 0) return null;
+  for (let i = packageStart + 1; i < lines.length; i++) {
+    const line = lines[i].trimStart();
+    if (line.startsWith('[')) break; // hit the next section header
+    const m = /^name\s*=\s*"([^"]+)"/.exec(line);
+    if (m) return m[1];
+  }
+  return null;
+}
+
 /**
  * Parse a Cargo.toml to extract the crate name and workspace dependency
  * names. Uses simple line-based parsing — no TOML library needed for
@@ -47,12 +73,9 @@ async function parseCrateManifest(
     return null;
   }
 
-  let name = '';
+  const name = parseCargoPackageName(content) ?? '';
   const workspaceDeps: string[] = [];
 
-  const nameMatch = content.match(/^\[package\]\s*\n(?:[^\[]*?\n)*?name\s*=\s*"([^"]+)"/m);
-  if (nameMatch) name = nameMatch[1];
-
   // Match dependencies that use workspace = true, which indicates they
   // are workspace-internal deps:
   //   dep_name = { workspace = true }
diff --git a/gitnexus/src/core/group/extractors/thrift-extractor.ts b/gitnexus/src/core/group/extractors/thrift-extractor.ts
index cfd8fef02..709968790 100644
--- a/gitnexus/src/core/group/extractors/thrift-extractor.ts
+++ b/gitnexus/src/core/group/extractors/thrift-extractor.ts
@@ -217,6 +217,10 @@ export async function buildThriftContext(repoPath: string): Promise<ThriftContex
     cwd: repoPath,
     absolute: false,
     nodir: true,
+    // TODO(#1156-followup): replace this hand-rolled list with createIgnoreFilter
+    // (the canonical ingestion ignore filter, like include-extractor.ts now uses).
+    // New entries to DEFAULT_IGNORE_LIST in src/config/ignore-service.ts (e.g.
+    // third_party, 3rdparty added in commit a9936a9b) silently do not apply here.
     ignore: ['**/node_modules/**', '**/.git/**', '**/vendor/**', '**/dist/**', '**/build/**'],
   });
   const namespacesByThrift = new Map<string, string>();
@@ -290,6 +294,10 @@ export class ThriftExtractor implements ContractExtractor {
       cwd: repoPath,
       absolute: false,
       nodir: true,
+      // TODO(#1156-followup): replace this hand-rolled list with createIgnoreFilter
+      // (the canonical ingestion ignore filter, like include-extractor.ts now uses).
+      // New entries to DEFAULT_IGNORE_LIST in src/config/ignore-service.ts (e.g.
+      // third_party, 3rdparty added in commit a9936a9b) silently do not apply here.
       ignore: ['**/node_modules/**', '**/.git/**', '**/vendor/**', '**/dist/**', '**/build/**'],
     });
 
diff --git a/gitnexus/src/core/group/matching.ts b/gitnexus/src/core/group/matching.ts
index 3431f8ddf..0b27655c6 100644
--- a/gitnexus/src/core/group/matching.ts
+++ b/gitnexus/src/core/group/matching.ts
@@ -107,6 +107,8 @@ export function normalizeContractId(id: string): string {
       return `topic::${rest.trim().toLowerCase()}`;
     case 'lib':
       return `lib::${rest.toLowerCase()}`;
+    case 'include':
+      return `include::${rest.replace(/\\/g, '/').replace(/^\.\//, '').replace(/\/+/g, '/').toLowerCase()}`;
     default:
       return id;
   }
diff --git a/gitnexus/src/core/group/service.ts b/gitnexus/src/core/group/service.ts
index de324e70b..d0473048f 100644
--- a/gitnexus/src/core/group/service.ts
+++ b/gitnexus/src/core/group/service.ts
@@ -65,6 +65,15 @@ export interface GroupToolPort {
       relationTypes: string[];
       minConfidence: number;
       includeTests: boolean;
+      // Optional cancellation signal. Callers (notably the cross-impact
+      // Phase-2 fanout) wrap this call in a Promise.race against a
+      // setTimeout-driven AbortController so a single hung neighbor
+      // cannot exceed the request's clamped timeout budget. Implementors
+      // may honor the signal cooperatively or simply let the caller's
+      // race resolve the await — the latter is sufficient for the
+      // resource-exhaustion mitigation. When the signal is absent or
+      // already aborted at call time, behavior is unchanged.
+      signal?: AbortSignal;
     },
   ): Promise<unknown | null>;
   context(
diff --git a/gitnexus/src/core/group/storage.ts b/gitnexus/src/core/group/storage.ts
index cc3dbfdc9..bc08fd7f9 100644
--- a/gitnexus/src/core/group/storage.ts
+++ b/gitnexus/src/core/group/storage.ts
@@ -4,6 +4,7 @@ import * as path from 'node:path';
 import * as os from 'node:os';
 import { randomBytes } from 'node:crypto';
 import type { ContractRegistry } from './types.js';
+import { retryRename } from './bridge-db.js';
 
 /**
  * Build an unpredictable suffix for atomic-write tmp files. Replaces the
@@ -59,7 +60,13 @@ export async function writeContractRegistry(
   } finally {
     await handle.close();
   }
-  await fsp.rename(tmpPath, targetPath);
+  // retryRename absorbs the documented Windows EPERM/EBUSY/EACCES race that
+  // fires when AV scanners or another concurrent rename briefly hold the
+  // destination handle between rename calls. Same helper bridge-db.ts uses
+  // (lines 304, 583, 587, 595, 605, 677) for the bridge.lbug atomic swap —
+  // single source of truth for the Windows-rename pattern across the group
+  // package.
+  await retryRename(tmpPath, targetPath);
 }
 
 export async function readContractRegistry(groupDir: string): Promise<ContractRegistry | null> {
diff --git a/gitnexus/src/core/group/sync.ts b/gitnexus/src/core/group/sync.ts
index 7ed065131..cd64fdf8c 100644
--- a/gitnexus/src/core/group/sync.ts
+++ b/gitnexus/src/core/group/sync.ts
@@ -8,12 +8,14 @@ import { HttpRouteExtractor } from './extractors/http-route-extractor.js';
 import { GrpcExtractor } from './extractors/grpc-extractor.js';
 import { ThriftExtractor } from './extractors/thrift-extractor.js';
 import { TopicExtractor } from './extractors/topic-extractor.js';
+import { IncludeExtractor } from './extractors/include-extractor.js';
 import { ManifestExtractor } from './extractors/manifest-extractor.js';
 import { discoverWorkspaceLinks } from './extractors/workspace-extractor.js';
 import { buildProviderIndex, runExactMatch, runWildcardMatch } from './matching.js';
 import { detectServiceBoundaries, assignService } from './service-boundary-detector.js';
 import type { CypherExecutor } from './contract-extractor.js';
 import { writeContractRegistry } from './storage.js';
+import { writeBridge } from './bridge-db.js';
 import type { ContractRegistry } from './types.js';
 
 import { logger } from '../logger.js';
@@ -100,6 +102,7 @@ export async function syncGroup(config: GroupConfig, opts?: SyncOptions): Promis
     const grpcEx = new GrpcExtractor();
     const thriftEx = new ThriftExtractor();
     const topicEx = new TopicExtractor();
+    const includeEx = new IncludeExtractor();
     dbExecutors = new Map<string, CypherExecutor>();
     const openPoolIds: string[] = [];
 
@@ -168,6 +171,17 @@ export async function syncGroup(config: GroupConfig, opts?: SyncOptions): Promis
             }
           }
 
+          if (config.detect.includes) {
+            const extracted = await includeEx.extract(executor, handle.repoPath, handle);
+            for (const c of extracted) {
+              autoContracts.push({
+                ...c,
+                repo: groupPath,
+                service: assignService(c.symbolRef.filePath, boundaries),
+              });
+            }
+          }
+
           const metaPath = path.join(handle.storagePath, 'meta.json');
           try {
             const raw = await fs.readFile(metaPath, 'utf-8');
@@ -270,6 +284,28 @@ export async function syncGroup(config: GroupConfig, opts?: SyncOptions): Promis
 
   if (opts?.groupDir && !opts.skipWrite) {
     await writeContractRegistry(opts.groupDir, registry);
+    // writeBridge failure (disk full, schema error, permission denied) must
+    // not mask the registry — contracts.json was just written successfully
+    // and is the canonical source of truth. A stale or absent bridge
+    // degrades impact queries to empty results, which is recoverable on
+    // the next sync. Surface the failure as a warning so operators can
+    // act, but do not propagate it.
+    // (PR #1156 follow-up review: writeBridge error in sync.ts propagates
+    // uncaught.)
+    try {
+      await writeBridge(opts.groupDir, {
+        contracts: allContracts,
+        crossLinks,
+        repoSnapshots,
+        missingRepos,
+      });
+    } catch (err) {
+      const msg = err instanceof Error ? err.message : String(err);
+      logger.warn(
+        { err: msg, groupDir: opts.groupDir },
+        '⚠️ writeBridge failed; contracts.json is intact but bridge.lbug is stale. Re-run `gitnexus group sync` to retry.',
+      );
+    }
   }
 
   return {
diff --git a/gitnexus/src/core/group/types.ts b/gitnexus/src/core/group/types.ts
index 7d0a14251..8e43ff78f 100644
--- a/gitnexus/src/core/group/types.ts
+++ b/gitnexus/src/core/group/types.ts
@@ -1,4 +1,4 @@
-export type ContractType = 'http' | 'grpc' | 'thrift' | 'topic' | 'lib' | 'custom';
+export type ContractType = 'http' | 'grpc' | 'thrift' | 'topic' | 'lib' | 'custom' | 'include';
 export type MatchType = 'exact' | 'manifest' | 'wildcard' | 'bm25' | 'embedding';
 export type ContractRole = 'provider' | 'consumer';
 
@@ -28,6 +28,7 @@ export interface DetectConfig {
   topics: boolean;
   shared_libs: boolean;
   embedding_fallback: boolean;
+  includes: boolean;
   workspace_deps: boolean;
 }
 
diff --git a/gitnexus/src/core/ingestion/call-processor.ts b/gitnexus/src/core/ingestion/call-processor.ts
index 6c59578b0..bf0206057 100644
--- a/gitnexus/src/core/ingestion/call-processor.ts
+++ b/gitnexus/src/core/ingestion/call-processor.ts
@@ -769,9 +769,10 @@ export const processCalls = async (
 
     let tree = astCache.get(file.path);
     if (!tree) {
+      const parseContent = provider.preprocessSource?.(file.content, file.path) ?? file.content;
       try {
-        tree = parser.parse(file.content, undefined, {
-          bufferSize: getTreeSitterBufferSize(file.content),
+        tree = parser.parse(parseContent, undefined, {
+          bufferSize: getTreeSitterBufferSize(parseContent),
         });
       } catch (parseError) {
         continue;
@@ -3280,9 +3281,10 @@ export const extractFetchCallsFromFiles = async (
 
     let tree = astCache.get(file.path);
     if (!tree) {
+      const parseContent = provider.preprocessSource?.(file.content, file.path) ?? file.content;
       try {
-        tree = parser.parse(file.content, undefined, {
-          bufferSize: getTreeSitterBufferSize(file.content),
+        tree = parser.parse(parseContent, undefined, {
+          bufferSize: getTreeSitterBufferSize(parseContent),
         });
       } catch {
         continue;
diff --git a/gitnexus/src/core/ingestion/cobol/cobol-preprocessor.ts b/gitnexus/src/core/ingestion/cobol/cobol-preprocessor.ts
index 23cedb99b..34be6bc03 100644
--- a/gitnexus/src/core/ingestion/cobol/cobol-preprocessor.ts
+++ b/gitnexus/src/core/ingestion/cobol/cobol-preprocessor.ts
@@ -369,9 +369,20 @@ const RE_USE_AFTER =
   /\bUSE\s+(?:AFTER\s+)?(?:STANDARD\s+)?(?:EXCEPTION|ERROR)\s+ON\s+([A-Z][A-Z0-9-]+|INPUT|OUTPUT|I-O|EXTEND)\b/i;
 
 // SET statement (condition, index)
-const RE_SET_TO_TRUE = /\bSET\s+((?:[A-Z][A-Z0-9-]+(?:\s+OF\s+[A-Z][A-Z0-9-]+)?\s+)+)TO\s+TRUE\b/i;
-const RE_SET_INDEX =
-  /\bSET\s+((?:[A-Z][A-Z0-9-]+\s+)+)(TO|UP\s+BY|DOWN\s+BY)\s+(\d+|[A-Z][A-Z0-9-]+)/i;
+//
+// Catastrophic-backtracking note (CodeQL js/redos): the previous shape
+// `((?:[A-Z][A-Z0-9-]+(?:\s+OF\s+[A-Z][A-Z0-9-]+)?\s+)+)TO\s+TRUE`
+// nested `\s+` quantifiers across alternations and was exponential on
+// inputs like "SET a OF a OF a ... TO TRUE". Replaced with a lazy
+// dot-match bounded by the explicit `\s+TO\s+TRUE` suffix — `.+?` is
+// O(n) with the trailing anchor, and the captured group is parsed
+// downstream the same way as before.
+// Exported so the U8 ReDoS regression test can pin the exact production
+// pattern. Direct import is the only way to ensure the test's
+// pathological-input timing assertion exercises the production regex
+// instead of an inline copy that drifts.
+export const RE_SET_TO_TRUE = /\bSET\s+(.+?)\s+TO\s+TRUE\b/i;
+export const RE_SET_INDEX = /\bSET\s+(.+?)\s+(TO|UP\s+BY|DOWN\s+BY)\s+(\d+|[A-Z][A-Z0-9-]+)/i;
 
 // INITIALIZE statement — data reset (captures targets before REPLACING/WITH clause)
 const RE_INITIALIZE = /\bINITIALIZE\s+([\s\S]*?)(?=\bREPLACING\b|\bWITH\b|\.\s*$|$)/i;
diff --git a/gitnexus/src/core/ingestion/cpp-ue-preprocessor.ts b/gitnexus/src/core/ingestion/cpp-ue-preprocessor.ts
new file mode 100644
index 000000000..baa5d17ff
--- /dev/null
+++ b/gitnexus/src/core/ingestion/cpp-ue-preprocessor.ts
@@ -0,0 +1,265 @@
+/**
+ * Unreal Engine reflection-macro preprocessor for C++ source.
+ *
+ * Tree-sitter does not expand C preprocessor macros, so Unreal's reflection
+ * markers (`UCLASS(...)`, `UFUNCTION(...)`, `MODULENAME_API`, ...) are parsed
+ * verbatim. The result is mis-parsed declarations: in `class BRAWLUI_API
+ * UMyClass : public UObject`, tree-sitter-cpp captures `BRAWLUI_API` as the
+ * class name and the rest of the declaration becomes structurally wrong.
+ *
+ * This module elides those macros from the source text BEFORE tree-sitter
+ * parses it. Replacement is **length-preserving** (each elided byte becomes
+ * a space, newlines preserved) so byte offsets and line/column positions
+ * tree-sitter reports remain identical to the original file. Symbol
+ * locations in the graph stay accurate.
+ *
+ * A cheap detection guard short-circuits files that don't look like UE
+ * sources, so non-UE C++ codebases pay no cost.
+ *
+ * Pure function — no tree-sitter dependency, safe for worker threads.
+ */
+/**
+ * Strong UE markers — reflection macros that only Unreal Engine projects use.
+ * Presence of one of these is sufficient evidence that the file is a UE source
+ * and that `MODULENAME_API` tokens in it are intended as export macros.
+ *
+ * Importantly, `_API` tokens are NOT in this guard — `REST_API`, `HTTP_API`,
+ * `MY_LIB_API` and similar identifiers appear in plenty of non-UE C++ codebases
+ * as constants/enums/parameter names. We must not erase them just because the
+ * file mentions an `_API` token.
+ */
+const HAS_UE_HINT =
+  /\b(?:UCLASS|UFUNCTION|UPROPERTY|USTRUCT|UENUM|UINTERFACE|GENERATED_BODY|GENERATED_[A-Z_]+_BODY|UE_DEPRECATED|DECLARE_(?:DYNAMIC_)?(?:MULTICAST_)?DELEGATE)/;
+
+const SIMPLE_MACROS_NO_ARGS: readonly string[] = [
+  'GENERATED_BODY',
+  'GENERATED_UCLASS_BODY',
+  'GENERATED_USTRUCT_BODY',
+  'GENERATED_UINTERFACE_BODY',
+  'GENERATED_IINTERFACE_BODY',
+  'DECLARE_CLASS',
+  'GENERATED_BODY_LEGACY',
+];
+
+const PARENTHESIZED_MACROS: readonly string[] = [
+  'UCLASS',
+  'UFUNCTION',
+  'UPROPERTY',
+  'USTRUCT',
+  'UENUM',
+  'UINTERFACE',
+  'UMETA',
+  'UE_DEPRECATED',
+];
+
+const DELEGATE_MACRO_RE =
+  /\bDECLARE_(?:DYNAMIC_)?(?:MULTICAST_)?DELEGATE(?:_(?:RetVal_OneParam|RetVal_TwoParams|RetVal_ThreeParams|RetVal_FourParams|RetVal_FiveParams|RetVal_SixParams|RetVal_SevenParams|RetVal_EightParams|RetVal_NineParams|RetVal|OneParam|TwoParams|ThreeParams|FourParams|FiveParams|SixParams|SevenParams|EightParams|NineParams|TenParams))?(?=\s*\()/g;
+
+/**
+ * Module export tokens like `BRAWLUI_API`, `ENGINE_API`, `COREUOBJECT_API`.
+ * Pattern: ALL_CAPS identifier ending in `_API`. The leading word boundary
+ * (`\b`) prevents matching mid-identifier.
+ */
+const API_MACRO_RE = /\b[A-Z][A-Z0-9_]*_API\b/g;
+
+/** Replace `[start, end)` of `chars` with spaces, preserving newlines. */
+function eraseRange(chars: string[], start: number, end: number): void {
+  for (let i = start; i < end; i++) {
+    if (chars[i] !== '\n' && chars[i] !== '\r') {
+      chars[i] = ' ';
+    }
+  }
+}
+
+/**
+ * Find the matching close paren for an opening paren at index `openIdx`.
+ * Returns the index of `)` (inclusive end), or -1 if unbalanced.
+ *
+ * Handles nested parens and string/char literals so commas/parens inside
+ * strings don't throw off the match. Does not attempt to handle raw string
+ * literals (`R"(...)"`); UE reflection-macro arguments do not use them in
+ * practice.
+ */
+function findMatchingParen(source: string, openIdx: number): number {
+  if (source.charCodeAt(openIdx) !== 0x28) return -1;
+  let depth = 1;
+  let i = openIdx + 1;
+  const len = source.length;
+  while (i < len && depth > 0) {
+    const ch = source.charCodeAt(i);
+    // String literal
+    if (ch === 0x22) {
+      i++;
+      while (i < len) {
+        const c = source.charCodeAt(i);
+        if (c === 0x5c) {
+          i += 2;
+          continue;
+        }
+        if (c === 0x22) {
+          i++;
+          break;
+        }
+        i++;
+      }
+      continue;
+    }
+    // Char literal
+    if (ch === 0x27) {
+      i++;
+      while (i < len) {
+        const c = source.charCodeAt(i);
+        if (c === 0x5c) {
+          i += 2;
+          continue;
+        }
+        if (c === 0x27) {
+          i++;
+          break;
+        }
+        i++;
+      }
+      continue;
+    }
+    // Line comment
+    if (ch === 0x2f && source.charCodeAt(i + 1) === 0x2f) {
+      while (i < len && source.charCodeAt(i) !== 0x0a) i++;
+      continue;
+    }
+    // Block comment
+    if (ch === 0x2f && source.charCodeAt(i + 1) === 0x2a) {
+      i += 2;
+      while (i < len) {
+        if (source.charCodeAt(i) === 0x2a && source.charCodeAt(i + 1) === 0x2f) {
+          i += 2;
+          break;
+        }
+        i++;
+      }
+      continue;
+    }
+    if (ch === 0x28) depth++;
+    else if (ch === 0x29) {
+      depth--;
+      if (depth === 0) return i;
+    }
+    i++;
+  }
+  return -1;
+}
+
+/** Match a whole-word identifier at `idx`. Returns the byte after the identifier, or -1 on miss. */
+function matchIdentifierAt(source: string, idx: number, name: string): number {
+  if (idx > 0) {
+    const prev = source.charCodeAt(idx - 1);
+    if (
+      (prev >= 0x30 && prev <= 0x39) ||
+      (prev >= 0x41 && prev <= 0x5a) ||
+      (prev >= 0x61 && prev <= 0x7a) ||
+      prev === 0x5f
+    ) {
+      return -1;
+    }
+  }
+  for (let k = 0; k < name.length; k++) {
+    if (source.charCodeAt(idx + k) !== name.charCodeAt(k)) return -1;
+  }
+  const after = idx + name.length;
+  if (after < source.length) {
+    const next = source.charCodeAt(after);
+    if (
+      (next >= 0x30 && next <= 0x39) ||
+      (next >= 0x41 && next <= 0x5a) ||
+      (next >= 0x61 && next <= 0x7a) ||
+      next === 0x5f
+    ) {
+      return -1;
+    }
+  }
+  return after;
+}
+
+/** Skip ASCII whitespace forward from `idx`. Returns the next non-whitespace byte index. */
+function skipWhitespace(source: string, idx: number): number {
+  const len = source.length;
+  while (idx < len) {
+    const ch = source.charCodeAt(idx);
+    if (ch === 0x20 || ch === 0x09 || ch === 0x0a || ch === 0x0d) {
+      idx++;
+      continue;
+    }
+    break;
+  }
+  return idx;
+}
+
+/**
+ * Strip Unreal Engine reflection macros from C++ source, length-preserving.
+ *
+ * Returns the original string unchanged if no strong UE marker is detected,
+ * so non-UE C++ files (including ones that contain `*_API`-suffixed
+ * identifiers like `REST_API` or `HTTP_API`) incur only a single regex test.
+ *
+ * The `_filePath` parameter is part of the `LanguageProvider.preprocessSource`
+ * contract but is unused — UE detection is purely content-based. Accepted and
+ * ignored here so the function matches the hook signature exactly.
+ */
+export function stripUeMacros(source: string, _filePath?: string): string {
+  if (!HAS_UE_HINT.test(source)) return source;
+
+  const chars: string[] = source.split('');
+
+  for (const macro of PARENTHESIZED_MACROS) {
+    let searchFrom = 0;
+    while (true) {
+      const hit = source.indexOf(macro, searchFrom);
+      if (hit < 0) break;
+      searchFrom = hit + 1;
+      const after = matchIdentifierAt(source, hit, macro);
+      if (after < 0) continue;
+      const parenIdx = skipWhitespace(source, after);
+      if (source.charCodeAt(parenIdx) !== 0x28) continue;
+      const close = findMatchingParen(source, parenIdx);
+      if (close < 0) continue;
+      eraseRange(chars, hit, close + 1);
+    }
+  }
+
+  for (const macro of SIMPLE_MACROS_NO_ARGS) {
+    let searchFrom = 0;
+    while (true) {
+      const hit = source.indexOf(macro, searchFrom);
+      if (hit < 0) break;
+      searchFrom = hit + 1;
+      const after = matchIdentifierAt(source, hit, macro);
+      if (after < 0) continue;
+      const parenIdx = skipWhitespace(source, after);
+      if (source.charCodeAt(parenIdx) === 0x28) {
+        const close = findMatchingParen(source, parenIdx);
+        if (close < 0) continue;
+        eraseRange(chars, hit, close + 1);
+      } else {
+        eraseRange(chars, hit, after);
+      }
+    }
+  }
+
+  for (const re of [DELEGATE_MACRO_RE, API_MACRO_RE]) {
+    re.lastIndex = 0;
+    let match: RegExpExecArray | null;
+    while ((match = re.exec(source)) !== null) {
+      const start = match.index;
+      let end = start + match[0].length;
+      if (re === DELEGATE_MACRO_RE) {
+        const parenIdx = skipWhitespace(source, end);
+        if (source.charCodeAt(parenIdx) === 0x28) {
+          const close = findMatchingParen(source, parenIdx);
+          if (close >= 0) end = close + 1;
+        }
+      }
+      eraseRange(chars, start, end);
+    }
+  }
+
+  return chars.join('');
+}
diff --git a/gitnexus/src/core/ingestion/field-extractors/configs/csharp.ts b/gitnexus/src/core/ingestion/field-extractors/configs/csharp.ts
index 4d83a6cd7..75032a9d7 100644
--- a/gitnexus/src/core/ingestion/field-extractors/configs/csharp.ts
+++ b/gitnexus/src/core/ingestion/field-extractors/configs/csharp.ts
@@ -9,6 +9,11 @@ import type { SyntaxNode } from '../../utils/ast-helpers.js';
 
 const CSHARP_VIS = new Set<FieldVisibility>(['public', 'private', 'protected', 'internal']);
 
+const extractCsharpDeclaredType = (typeNode: SyntaxNode): string | undefined => {
+  if (typeNode.type === 'generic_name') return typeNode.text.trim();
+  return extractSimpleTypeName(typeNode) ?? typeNode.text?.trim();
+};
+
 /**
  * C# field extraction config.
  *
@@ -53,17 +58,17 @@ export const csharpConfig: FieldExtractionConfig = {
       const child = node.namedChild(i);
       if (child?.type === 'variable_declaration') {
         const typeNode = child.childForFieldName('type');
-        if (typeNode) return extractSimpleTypeName(typeNode) ?? typeNode.text?.trim();
+        if (typeNode) return extractCsharpDeclaredType(typeNode);
         // fallback: first child that is a type
         const first = child.firstNamedChild;
         if (first && first.type !== 'variable_declarator') {
-          return extractSimpleTypeName(first) ?? first.text?.trim();
+          return extractCsharpDeclaredType(first);
         }
       }
     }
     // property_declaration: type is first named child
     const typeNode = node.childForFieldName('type');
-    if (typeNode) return extractSimpleTypeName(typeNode) ?? typeNode.text?.trim();
+    if (typeNode) return extractCsharpDeclaredType(typeNode);
     return undefined;
   },
 
diff --git a/gitnexus/src/core/ingestion/heritage-processor.ts b/gitnexus/src/core/ingestion/heritage-processor.ts
index 2c973ad8e..f8628e651 100644
--- a/gitnexus/src/core/ingestion/heritage-processor.ts
+++ b/gitnexus/src/core/ingestion/heritage-processor.ts
@@ -219,9 +219,13 @@ export const processHeritage = async (
     let tree = astCache.get(file.path);
     if (!tree) {
       // Use larger bufferSize for files > 32KB
+      // Per-language source preprocessor (length-preserving, e.g. UE macro
+      // stripping for C++). MUST mirror parsing-processor on cache miss so
+      // re-parses see the same input as the cached AST.
+      const parseContent = provider.preprocessSource?.(file.content, file.path) ?? file.content;
       try {
-        tree = parser.parse(file.content, undefined, {
-          bufferSize: getTreeSitterBufferSize(file.content),
+        tree = parser.parse(parseContent, undefined, {
+          bufferSize: getTreeSitterBufferSize(parseContent),
         });
       } catch (parseError) {
         // Skip files that can't be parsed
@@ -413,9 +417,10 @@ export async function extractExtractedHeritageFromFiles(
 
     let tree = astCache.get(file.path);
     if (!tree) {
+      const parseContent = provider.preprocessSource?.(file.content, file.path) ?? file.content;
       try {
-        tree = parser.parse(file.content, undefined, {
-          bufferSize: getTreeSitterBufferSize(file.content),
+        tree = parser.parse(parseContent, undefined, {
+          bufferSize: getTreeSitterBufferSize(parseContent),
         });
       } catch {
         continue;
diff --git a/gitnexus/src/core/ingestion/import-processor.ts b/gitnexus/src/core/ingestion/import-processor.ts
index 03cbed5b7..6f0b40b6f 100644
--- a/gitnexus/src/core/ingestion/import-processor.ts
+++ b/gitnexus/src/core/ingestion/import-processor.ts
@@ -305,9 +305,10 @@ export const processImports = async (
     let wasReparsed = false;
 
     if (!tree) {
+      const parseContent = provider.preprocessSource?.(file.content, file.path) ?? file.content;
       try {
-        tree = parser.parse(file.content, undefined, {
-          bufferSize: getTreeSitterBufferSize(file.content),
+        tree = parser.parse(parseContent, undefined, {
+          bufferSize: getTreeSitterBufferSize(parseContent),
         });
       } catch (parseError) {
         continue;
diff --git a/gitnexus/src/core/ingestion/language-provider.ts b/gitnexus/src/core/ingestion/language-provider.ts
index 12ce0839b..e139cf5f3 100644
--- a/gitnexus/src/core/ingestion/language-provider.ts
+++ b/gitnexus/src/core/ingestion/language-provider.ts
@@ -116,6 +116,39 @@ interface LanguageProviderConfig {
    *  Required for tree-sitter languages; empty string for standalone processors. */
   readonly treeSitterQueries: string;
 
+  /**
+   * Optional source-text transform that runs **before** tree-sitter parses the file.
+   *
+   * Used to elide language constructs that confuse the grammar without affecting
+   * source-position fidelity — e.g., Unreal Engine reflection macros (`UCLASS`,
+   * `UFUNCTION`, `MODULENAME_API`) in C++ headers that prevent the parser from
+   * recognising class/function names correctly.
+   *
+   * **Length / position preservation:** the returned string MUST have the same
+   * JavaScript `.length` as the input AND preserve every newline (`\n`/`\r`)
+   * position byte-for-byte. Implementations replace elided characters with
+   * ASCII spaces while leaving newlines untouched. With this contract:
+   *
+   *   - tree-sitter's reported `startPosition.row`/`startPosition.column`
+   *     match the original file exactly (line/column come from newline counts)
+   *   - `startIndex`/`endIndex` byte offsets match the original file exactly
+   *     **when the elided range is pure ASCII** (UTF-16 `.length` equals UTF-8
+   *     byte length only for ASCII).
+   *
+   * Implementations targeting languages where elided ranges may contain
+   * non-ASCII content must therefore preserve byte length, not just `.length`,
+   * if downstream code uses `startIndex` to slice the original UTF-8 bytes.
+   * The current C++ UE-macro preprocessor relies on the practical fact that
+   * UE reflection macros and module-export tokens are ASCII-only.
+   *
+   * Must be a pure function — same input always yields the same output. Called
+   * once per file, on every code path that re-parses (parsing-processor, import
+   * processor, heritage processor, call processor, parse worker).
+   *
+   * Default: undefined (no preprocessing — `file.content` is parsed verbatim).
+   */
+  readonly preprocessSource?: (sourceText: string, filePath: string) => string;
+
   // ── Core (required) ───────────────────────────────────────────────
   /** Type extraction: declarations, initializers, for-loop bindings */
   readonly typeConfig: LanguageTypeConfig;
diff --git a/gitnexus/src/core/ingestion/languages/c-cpp.ts b/gitnexus/src/core/ingestion/languages/c-cpp.ts
index a5b3e5729..693e8cec2 100644
--- a/gitnexus/src/core/ingestion/languages/c-cpp.ts
+++ b/gitnexus/src/core/ingestion/languages/c-cpp.ts
@@ -45,6 +45,7 @@ import { cVariableConfig, cppVariableConfig } from '../variable-extractors/confi
 import { createCallExtractor } from '../call-extractors/generic.js';
 import { cCallConfig, cppCallConfig } from '../call-extractors/configs/c-cpp.js';
 import { createHeritageExtractor } from '../heritage-extractors/generic.js';
+import { stripUeMacros } from '../cpp-ue-preprocessor.js';
 
 const C_BUILT_INS: ReadonlySet<string> = new Set([
   'printf',
@@ -410,6 +411,7 @@ export const cppProvider = defineLanguage({
     },
   ] satisfies AstFrameworkPatternConfig[],
   treeSitterQueries: CPP_QUERIES,
+  preprocessSource: stripUeMacros,
   typeConfig: cCppConfig,
   exportChecker: cCppExportChecker,
   importResolver: createImportResolver(cppImportConfig),
diff --git a/gitnexus/src/core/ingestion/languages/csharp/captures.ts b/gitnexus/src/core/ingestion/languages/csharp/captures.ts
index 2f913a14e..e29552e78 100644
--- a/gitnexus/src/core/ingestion/languages/csharp/captures.ts
+++ b/gitnexus/src/core/ingestion/languages/csharp/captures.ts
@@ -42,6 +42,39 @@ const FUNCTION_NODE_TYPES = [
   'local_function_statement',
 ] as const;
 
+const BUILTIN_TYPE_NAMES = new Set([
+  'bool',
+  'byte',
+  'char',
+  'decimal',
+  'double',
+  'float',
+  'int',
+  'long',
+  'object',
+  'sbyte',
+  'short',
+  'string',
+  'uint',
+  'ulong',
+  'ushort',
+  'void',
+]);
+
+function shouldEmitReadMember(memberNode: SyntaxNode): boolean {
+  const parent = memberNode.parent;
+  if (parent === null) return true;
+
+  switch (parent.type) {
+    case 'invocation_expression':
+      return parent.childForFieldName('function')?.id !== memberNode.id;
+    case 'assignment_expression':
+      return parent.childForFieldName('left')?.id !== memberNode.id;
+    default:
+      return true;
+  }
+}
+
 export function emitCsharpScopeCaptures(
   sourceText: string,
   _filePath: string,
@@ -94,6 +127,14 @@ export function emitCsharpScopeCaptures(
       continue;
     }
 
+    if (grouped['@reference.read.member'] !== undefined) {
+      const anchor = grouped['@reference.read.member'];
+      const memberNode = findNodeAtRange(tree.rootNode, anchor.range, 'member_access_expression');
+      if (memberNode === null || !shouldEmitReadMember(memberNode)) {
+        continue;
+      }
+    }
+
     // Synthesize `this` / `base` receiver type-bindings on every
     // instance method-like. Tree-sitter can't cleanly express "the
     // implicit receiver of a non-static member of a class/struct/
@@ -209,9 +250,63 @@ export function emitCsharpScopeCaptures(
     }
   }
 
+  out.push(...synthesizeGenericTypeArgumentReferences(tree.rootNode));
+
   return out;
 }
 
+function synthesizeGenericTypeArgumentReferences(root: SyntaxNode): CaptureMatch[] {
+  const out: CaptureMatch[] = [];
+  // Treat all generic type arguments as static type references, including
+  // declaration signatures and call-site generic instantiations.
+  visit(root, (node) => {
+    if (node.type !== 'generic_name') return;
+    const args = findNamedChild(node, 'type_argument_list');
+    if (args === null) return;
+
+    for (const arg of args.namedChildren) {
+      if (arg === null) continue;
+      const nameNode = terminalTypeNameNode(arg);
+      if (nameNode === null) continue;
+      if (BUILTIN_TYPE_NAMES.has(nameNode.text)) continue;
+      out.push({
+        '@reference.type': nodeToCapture('@reference.type', nameNode),
+        '@reference.name': nodeToCapture('@reference.name', nameNode),
+      });
+    }
+  });
+  return out;
+}
+
+function terminalTypeNameNode(node: SyntaxNode): SyntaxNode | null {
+  switch (node.type) {
+    case 'identifier':
+      return node;
+    case 'nullable_type':
+      return node.firstNamedChild === null ? null : terminalTypeNameNode(node.firstNamedChild);
+    case 'qualified_name':
+      return node.lastNamedChild;
+    case 'generic_name':
+      return node.childForFieldName('name') ?? node.firstNamedChild;
+    default:
+      return null;
+  }
+}
+
+function findNamedChild(node: SyntaxNode, type: string): SyntaxNode | null {
+  for (const child of node.namedChildren) {
+    if (child !== null && child.type === type) return child;
+  }
+  return null;
+}
+
+function visit(node: SyntaxNode, cb: (node: SyntaxNode) => void): void {
+  cb(node);
+  for (const child of node.namedChildren) {
+    if (child !== null) visit(child, cb);
+  }
+}
+
 /** C# 12 primary constructor: `class X(a, b) { }` / `record X(a, b)`.
  *  The parameters are a bare `parameter_list` named child of the type
  *  declaration (no `constructor_declaration` node). Emit a synthetic
diff --git a/gitnexus/src/core/ingestion/languages/csharp/query.ts b/gitnexus/src/core/ingestion/languages/csharp/query.ts
index 11da2d8cd..f299aaafa 100644
--- a/gitnexus/src/core/ingestion/languages/csharp/query.ts
+++ b/gitnexus/src/core/ingestion/languages/csharp/query.ts
@@ -499,6 +499,12 @@ const CSHARP_SCOPE_QUERY = `
   left: (member_access_expression
     expression: "base" @reference.receiver
     name: (identifier) @reference.name)) @reference.write.member
+
+;; References — field/property reads: \`obj.Name\`
+;; Emit-side filtering drops call targets and assignment left-hand sides.
+(member_access_expression
+  expression: (_) @reference.receiver
+  name: (identifier) @reference.name) @reference.read.member
 `;
 
 let _parser: Parser | null = null;
diff --git a/gitnexus/src/core/ingestion/parsing-processor.ts b/gitnexus/src/core/ingestion/parsing-processor.ts
index 8803ec023..98036fbe8 100644
--- a/gitnexus/src/core/ingestion/parsing-processor.ts
+++ b/gitnexus/src/core/ingestion/parsing-processor.ts
@@ -371,6 +371,11 @@ const processParsingSequential = async (
       isVueSetup = extracted.isSetup;
     }
 
+    // Per-language source-text transform (e.g., UE macro stripping for C++).
+    // Length-preserving — see LanguageProvider.preprocessSource contract.
+    parseContent =
+      getProvider(language).preprocessSource?.(parseContent, file.path) ?? parseContent;
+
     try {
       await loadLanguage(language, file.path);
     } catch {
diff --git a/gitnexus/src/core/ingestion/workers/parse-worker.ts b/gitnexus/src/core/ingestion/workers/parse-worker.ts
index 5c6712562..4442b5a65 100644
--- a/gitnexus/src/core/ingestion/workers/parse-worker.ts
+++ b/gitnexus/src/core/ingestion/workers/parse-worker.ts
@@ -1407,6 +1407,11 @@ const processFileGroup = (
       isVueSetup = extracted.isSetup;
     }
 
+    // Per-language source-text transform (e.g., UE macro stripping for C++).
+    // Length-preserving — see LanguageProvider.preprocessSource contract.
+    parseContent =
+      getProvider(language).preprocessSource?.(parseContent, file.path) ?? parseContent;
+
     clearCaches(); // Reset memoization before each new file
 
     let tree;
diff --git a/gitnexus/src/core/lbug/csv-generator.ts b/gitnexus/src/core/lbug/csv-generator.ts
index 616a00cac..20df51ebd 100644
--- a/gitnexus/src/core/lbug/csv-generator.ts
+++ b/gitnexus/src/core/lbug/csv-generator.ts
@@ -301,11 +301,15 @@ export const streamAllCSVsToDisk = async (
     'Template',
     'Module',
   ] as const;
+  const propertyHeader = 'id,name,filePath,startLine,endLine,content,description,declaredType';
   const multiLangWriters = new Map<string, BufferedCSVWriter>();
   for (const t of MULTI_LANG_TYPES) {
     multiLangWriters.set(
       t,
-      new BufferedCSVWriter(path.join(csvDir, `${t.toLowerCase()}.csv`), multiLangHeader),
+      new BufferedCSVWriter(
+        path.join(csvDir, `${t.toLowerCase()}.csv`),
+        t === 'Property' ? propertyHeader : multiLangHeader,
+      ),
     );
   }
 
@@ -478,6 +482,9 @@ export const streamAllCSVsToDisk = async (
                 escapeCSVNumber(node.properties.endLine, -1),
                 escapeCSVField(content),
                 escapeCSVField(node.properties.description || ''),
+                ...(node.label === 'Property'
+                  ? [escapeCSVField(node.properties.declaredType || '')]
+                  : []),
               ].join(','),
             );
           }
diff --git a/gitnexus/src/core/lbug/lbug-adapter.ts b/gitnexus/src/core/lbug/lbug-adapter.ts
index 2fc12cf96..cb88986ca 100644
--- a/gitnexus/src/core/lbug/lbug-adapter.ts
+++ b/gitnexus/src/core/lbug/lbug-adapter.ts
@@ -19,7 +19,10 @@ import type { CachedEmbedding } from '../embeddings/types.js';
 import { extensionManager, type ExtensionEnsureOptions } from './extension-loader.js';
 import {
   closeLbugConnection,
+  isDbBusyError,
+  isOpenRetryExhausted,
   openLbugConnection,
+  waitForWindowsHandleRelease,
   type LbugConnectionHandle,
 } from './lbug-config.js';
 import { isVectorExtensionSupportedByPlatform } from '../platform/capabilities.js';
@@ -185,21 +188,6 @@ const DB_LOCK_RETRY_ATTEMPTS = 3;
 /** Base back-off in ms between BUSY retries (multiplied by attempt number). */
 const DB_LOCK_RETRY_DELAY_MS = 500;
 
-/**
- * Return true when the error message indicates that another process holds
- * an exclusive lock on the LadybugDB file (e.g. `gitnexus analyze` or
- * `gitnexus serve` running at the same time).
- */
-export const isDbBusyError = (err: unknown): boolean => {
-  const msg = (err instanceof Error ? err.message : String(err)).toLowerCase();
-  return (
-    msg.includes('busy') ||
-    msg.includes('lock') ||
-    msg.includes('already in use') ||
-    msg.includes('could not set lock')
-  );
-};
-
 /**
  * Return true when the error message indicates a write was attempted against
  * a read-only LadybugDB connection. The MCP query pool opens DBs read-only,
@@ -252,7 +240,11 @@ export const withLbugDb = async <T>(dbPath: string, operation: () => Promise<T>)
       });
     } catch (err) {
       lastError = err;
-      if (!isDbBusyError(err) || attempt === DB_LOCK_RETRY_ATTEMPTS) {
+      // Skip outer retry when the inner open-retry already exhausted: the
+      // ~1.5s open-time budget was just spent, repeating the full reset+
+      // reopen cycle would only add 4-5s of tail latency without changing
+      // the outcome (both layers consult the same isDbBusyError matcher).
+      if (!isDbBusyError(err) || isOpenRetryExhausted(err) || attempt === DB_LOCK_RETRY_ATTEMPTS) {
         throw err;
       }
       // Close stale connection inside the session lock to prevent race conditions
@@ -330,7 +322,16 @@ const doInitLbug = async (dbPath: string) => {
       await conn.query(schemaQuery);
     } catch (err) {
       const msg = err instanceof Error ? err.message : String(err);
-      if (!msg.includes('already exists')) {
+      // Suppression list:
+      //   - "already exists": expected idempotent re-create on existing DBs
+      //   - "could not set lock on file": LadybugDB v0.16.1 emits this on
+      //     Windows when CREATE NODE TABLE runs against a path that was
+      //     just opened (the WAL handle from a fresh Database briefly
+      //     contests the table's first-write lock). The table is created
+      //     anyway and any genuine cross-process lock contention surfaces
+      //     on the next operation via withLbugDb's retry. Logging it here
+      //     would just be noise in CI.
+      if (!msg.includes('already exists') && !isDbBusyError(err)) {
         logger.warn(`⚠️ Schema creation warning: ${msg.slice(0, 120)}`);
       }
     }
@@ -607,6 +608,9 @@ const getCopyQuery = (table: NodeTableName, filePath: string): string => {
   if (table === 'Method') {
     return `COPY ${t}(id, name, filePath, startLine, endLine, isExported, content, description, parameterCount, returnType) FROM "${filePath}" ${COPY_CSV_OPTS}`;
   }
+  if (table === 'Property') {
+    return `COPY ${t}(id, name, filePath, startLine, endLine, content, description, declaredType) FROM "${filePath}" ${COPY_CSV_OPTS}`;
+  }
   // TypeScript/JS code element tables have isExported; multi-language tables do not
   if (TABLES_WITH_EXPORTED.has(table)) {
     return `COPY ${t}(id, name, filePath, startLine, endLine, isExported, content, description) FROM "${filePath}" ${COPY_CSV_OPTS}`;
@@ -658,6 +662,11 @@ export const insertNodeToLbug = async (
         ? `, description: ${escapeValue(properties.description)}`
         : '';
       query = `CREATE (n:${t} {id: ${escapeValue(properties.id)}, name: ${escapeValue(properties.name)}, filePath: ${escapeValue(properties.filePath)}, startLine: ${properties.startLine || 0}, endLine: ${properties.endLine || 0}, isExported: ${!!properties.isExported}, content: ${escapeValue(properties.content || '')}${descPart}})`;
+    } else if (label === 'Property') {
+      const descPart = properties.description
+        ? `, description: ${escapeValue(properties.description)}`
+        : '';
+      query = `CREATE (n:${t} {id: ${escapeValue(properties.id)}, name: ${escapeValue(properties.name)}, filePath: ${escapeValue(properties.filePath)}, startLine: ${properties.startLine || 0}, endLine: ${properties.endLine || 0}, content: ${escapeValue(properties.content || '')}${descPart}, declaredType: ${escapeValue(properties.declaredType || '')}})`;
     } else {
       // Multi-language tables (Struct, Impl, Trait, Macro, etc.) — no isExported
       const descPart = properties.description
@@ -736,6 +745,11 @@ export const batchInsertNodesToLbug = async (
             ? `, n.description = ${escapeValue(properties.description)}`
             : '';
           query = `MERGE (n:${t} {id: ${escapeValue(properties.id)}}) SET n.name = ${escapeValue(properties.name)}, n.filePath = ${escapeValue(properties.filePath)}, n.startLine = ${properties.startLine || 0}, n.endLine = ${properties.endLine || 0}, n.isExported = ${!!properties.isExported}, n.content = ${escapeValue(properties.content || '')}${descPart}`;
+        } else if (label === 'Property') {
+          const descPart = properties.description
+            ? `, n.description = ${escapeValue(properties.description)}`
+            : '';
+          query = `MERGE (n:${t} {id: ${escapeValue(properties.id)}}) SET n.name = ${escapeValue(properties.name)}, n.filePath = ${escapeValue(properties.filePath)}, n.startLine = ${properties.startLine || 0}, n.endLine = ${properties.endLine || 0}, n.content = ${escapeValue(properties.content || '')}${descPart}, n.declaredType = ${escapeValue(properties.declaredType || '')}`;
         } else {
           const descPart = properties.description
             ? `, n.description = ${escapeValue(properties.description)}`
@@ -1064,6 +1078,9 @@ export const flushWAL = async (): Promise<void> => {
  */
 export const safeClose = async (): Promise<void> => {
   await flushWAL();
+  // Capture before close — currentDbPath stays set so the Windows post-close
+  // probe below knows which file to wait on.
+  const closingDbPath = currentDbPath;
   if (conn) {
     try {
       // eslint-disable-next-line no-restricted-syntax -- sole authorised close site
@@ -1082,6 +1099,24 @@ export const safeClose = async (): Promise<void> => {
     }
     db = null;
   }
+  // Windows: libuv reports `db.close()` resolved before the kernel has
+  // released the file handle. A subsequent `new Database(samePath)` in
+  // the same process can race the release. The probe (lbug-config.ts)
+  // forces any residual lock to surface as EBUSY/EPERM/EACCES so the
+  // open-time retry absorbs the lag.
+  if (process.platform === 'win32' && closingDbPath) {
+    const released = await waitForWindowsHandleRelease(closingDbPath);
+    if (!released) {
+      // Probe exhausted with a lock code still in flight. The next
+      // openLbugConnection will absorb whatever residual lag remains, but
+      // a chronic warning helps operators spot AV interference (Windows
+      // Defender holding the file far past the 250ms budget).
+      logger.warn(
+        { dbPath: closingDbPath },
+        '⚠️ LadybugDB file handle still locked after close (Windows). If this repeats, check antivirus/Defender exclusions for the GitNexus storage directory.',
+      );
+    }
+  }
 };
 
 export const closeLbug = async (): Promise<void> => {
diff --git a/gitnexus/src/core/lbug/lbug-config.ts b/gitnexus/src/core/lbug/lbug-config.ts
index 5534f7d8b..ceb445693 100644
--- a/gitnexus/src/core/lbug/lbug-config.ts
+++ b/gitnexus/src/core/lbug/lbug-config.ts
@@ -1,3 +1,6 @@
+import fs from 'fs/promises';
+import os from 'os';
+import path from 'path';
 import type lbug from '@ladybugdb/core';
 
 /**
@@ -42,10 +45,23 @@ export const LBUG_MAX_DB_SIZE: number = (() => {
   return 16 * 1024 * 1024 * 1024;
 })();
 
+/** Matches WAL corruption errors from the LadybugDB engine. */
+const WAL_CORRUPTION_RE = /corrupt(ed)?\s+wal|invalid\s+wal\s+record|wal.*corrupt|checksum.*wal/i;
+
+export const WAL_RECOVERY_SUGGESTION =
+  'WAL corruption detected. Run `gitnexus analyze` to rebuild the index.';
+
+export function isWalCorruptionError(err: unknown): boolean {
+  if (!err) return false;
+  const msg = err instanceof Error ? err.message : String(err);
+  return WAL_CORRUPTION_RE.test(msg);
+}
+
 type LbugModule = typeof lbug;
 
 export interface LbugDatabaseOptions {
   readOnly?: boolean;
+  throwOnWalReplayFailure?: boolean;
 }
 
 export interface LbugConnectionHandle {
@@ -53,20 +69,200 @@ export interface LbugConnectionHandle {
   conn: lbug.Connection;
 }
 
+/**
+ * Return true when the error message indicates that a LadybugDB file lock
+ * could not be acquired — either at construction time
+ * (`new lbug.Database(...)` raises from `local_file_system.cpp`) or during
+ * a query (another writer holds the exclusive lock).
+ *
+ * Lives here (not in `lbug-adapter.ts`) so both the construction-time
+ * retry (`openWithLockRetry` in this file) and the query-time retry
+ * (`withLbugDb` in `lbug-adapter.ts`) consult the same matcher. Callers
+ * import directly from this module — no re-export to keep in sync.
+ */
+export const isDbBusyError = (err: unknown): boolean => {
+  const msg = (err instanceof Error ? err.message : String(err)).toLowerCase();
+  // `lock` already subsumes `could not set lock`; the broader term is kept
+  // because graph-DB transient errors include "deadlock", "lock contention",
+  // and the LadybugDB native module's "could not set lock on file" — all of
+  // which deserve a retry. If a non-transient lock-shaped error ever
+  // surfaces (e.g., "lock file missing" during recovery), tighten this
+  // matcher rather than raising the retry budget.
+  return msg.includes('busy') || msg.includes('lock') || msg.includes('already in use');
+};
+
 export function createLbugDatabase(
   lbugModule: LbugModule,
   databasePath: string,
   options: LbugDatabaseOptions = {},
 ): lbug.Database {
-  return new lbugModule.Database(
+  // .d.ts declares fewer args than the native constructor accepts.
+  return new (lbugModule.Database as any)(
     databasePath,
-    0,
-    false,
+    0, // bufferManagerSize
+    false, // enableCompression (pinned for v0.16.0)
     options.readOnly ?? false,
     LBUG_MAX_DB_SIZE,
-  );
+    true, // autoCheckpoint
+    -1, // checkpointThreshold
+    options.throwOnWalReplayFailure ?? true,
+    true, // enableChecksums
+  ) as lbug.Database;
 }
 
+// ─── Lock-busy retry tuning knobs ───────────────────────────────────────────
+//
+// All four GitNexus retry pairs that touch native LadybugDB locks live with
+// a comment cross-reference here so an SRE tuning Windows flakes finds them
+// in one grep:
+//
+//   1. OPEN_LOCK_RETRY_ATTEMPTS / OPEN_LOCK_RETRY_DELAY_MS  (this file)
+//      → `new lbug.Database()` constructor lock failures
+//   2. HANDLE_RELEASE_PROBE_ATTEMPTS / HANDLE_RELEASE_PROBE_DELAY_MS  (this file)
+//      → post-close fs.open probe to absorb Windows handle-release lag
+//   3. DB_LOCK_RETRY_ATTEMPTS / DB_LOCK_RETRY_DELAY_MS  (lbug-adapter.ts withLbugDb)
+//      → query-time busy/lock retry around already-open connections
+//
+// `new lbug.Database()` calls into the native module which performs an
+// OS-level exclusive lock on `<dbPath>`. On Windows that lock can fail
+// for reasons specific to the OS (Defender briefly opens new files,
+// libuv handle release lags the JS-side close). 5 attempts × 100ms
+// linear back-off (max sleep 100+200+300+400 = 1s, plus 5 ctor RTTs
+// of 10–50ms each = ~1.0–1.2s worst case) clears the typical
+// AV-scanner hold without masking real cross-process conflicts.
+//
+// Source: https://github.com/LadybugDB/ladybug/blob/v0.16.1/src/common/file_system/local_file_system.cpp#L126
+const OPEN_LOCK_RETRY_ATTEMPTS = 5;
+const OPEN_LOCK_RETRY_DELAY_MS = 100;
+
+const HANDLE_RELEASE_PROBE_ATTEMPTS = 5;
+const HANDLE_RELEASE_PROBE_DELAY_MS = 50;
+const HANDLE_RELEASE_LOCK_CODES = new Set(['EBUSY', 'EPERM', 'EACCES']);
+
+/**
+ * Test-fixture directory prefixes recognized by `isTestFixturePath`.
+ *
+ * IMPORTANT: this list must stay in sync with the prefixes passed to
+ * `createTempDir` in `gitnexus/test/helpers/test-db.ts` and the prefixes
+ * used by `withTestLbugDB` (`gitnexus/test/helpers/test-indexed-db.ts`).
+ * If you add a new test that passes a custom prefix to `createTempDir`,
+ * add it here too — otherwise the stale-sidecar sweep silently won't
+ * fire for that fixture and CI flakes return.
+ *
+ * The default `createTempDir('gitnexus-test-')` and the lbug variant
+ * `'gitnexus-lbug-'` cover today's call sites.
+ */
+const TEST_FIXTURE_PREFIXES = ['gitnexus-lbug-', 'gitnexus-test-'];
+
+/**
+ * Marker symbol attached to lock errors after `openWithLockRetry` exhausts
+ * its budget. `withLbugDb`'s outer query-time retry consults this so it
+ * does not re-retry a path that just spent up to ~1.5s in the open-time
+ * loop — preventing 6s tail latencies (3× outer × 5× inner attempts).
+ *
+ * The symbol is internal to GitNexus; consumers should treat the underlying
+ * error message as the user-visible signal.
+ */
+export const LBUG_OPEN_RETRY_EXHAUSTED = Symbol.for('gitnexus.lbug.openRetryExhausted');
+
+export const isOpenRetryExhausted = (err: unknown): boolean => {
+  if (err === null || err === undefined || typeof err !== 'object') return false;
+  return (err as { [LBUG_OPEN_RETRY_EXHAUSTED]?: boolean })[LBUG_OPEN_RETRY_EXHAUSTED] === true;
+};
+
+const tagOpenRetryExhausted = (err: unknown): unknown => {
+  if (err && typeof err === 'object') {
+    (err as { [LBUG_OPEN_RETRY_EXHAUSTED]?: boolean })[LBUG_OPEN_RETRY_EXHAUSTED] = true;
+  }
+  return err;
+};
+
+/**
+ * True when `dbPath` resolves to a recognized test fixture under the OS
+ * temp directory. Used to gate the stale-sidecar sweep so production
+ * paths never have their `.wal` / `.lock` files deleted.
+ *
+ * Defensive shape:
+ *   - `path.resolve` normalizes `..` segments before the prefix check, so
+ *     `<tmp>/gitnexus-lbug-x/../../etc/passwd` is rejected.
+ *   - The tmpRoot check trims any trailing separator returned by some
+ *     Windows TMP configurations (`C:\Users\X\Temp\`) so the startsWith
+ *     comparison stays correct.
+ *   - Only the IMMEDIATE parent directory is matched against the prefix
+ *     list. An ancestor walk would let a tmpdir whose own basename starts
+ *     with `gitnexus-lbug-` accept arbitrary nested paths under it.
+ */
+const isTestFixturePath = (dbPath: string): boolean => {
+  const tmpRoot = os.tmpdir().replace(new RegExp(`${path.sep === '\\' ? '\\\\' : path.sep}+$`), '');
+  const resolved = path.resolve(dbPath);
+  if (!resolved.startsWith(tmpRoot + path.sep) && resolved !== tmpRoot) return false;
+  const parentBase = path.basename(path.dirname(resolved));
+  return TEST_FIXTURE_PREFIXES.some((p) => parentBase.startsWith(p));
+};
+
+/** Exported only for direct unit testing — production callers use `openWithLockRetry`. */
+export const _isTestFixturePathForTest = isTestFixturePath;
+
+const sleep = (ms: number): Promise<void> => new Promise((resolve) => setTimeout(resolve, ms));
+
+/**
+ * Attempt to remove stale `.wal` / `.lock` sidecars that a previous aborted
+ * test run may have left behind. Best-effort: ENOENT is normal, anything
+ * else is swallowed so the caller's retry can surface the original error.
+ */
+const sweepStaleSidecars = async (dbPath: string): Promise<void> => {
+  for (const suffix of ['.wal', '.lock']) {
+    try {
+      await fs.unlink(dbPath + suffix);
+    } catch {
+      /* missing sidecar or permission error — let the open retry surface it */
+    }
+  }
+};
+
+/**
+ * Run `construct` with bounded retries when `new lbug.Database(...)` throws
+ * a busy/lock error. The original (loop-captured) error is preferred over
+ * any post-sweep error so triage sees the real LadybugDB lock message.
+ * On exhaustion the rethrown error is tagged via
+ * `LBUG_OPEN_RETRY_EXHAUSTED` so the outer query-time retry in
+ * `withLbugDb` skips re-retrying a freshly-exhausted path.
+ */
+const openWithLockRetry = async (
+  construct: () => lbug.Database,
+  dbPath: string,
+): Promise<lbug.Database> => {
+  let originalLockError: unknown;
+  for (let attempt = 1; attempt <= OPEN_LOCK_RETRY_ATTEMPTS; attempt++) {
+    try {
+      return construct();
+    } catch (err) {
+      if (!isDbBusyError(err)) throw err;
+      originalLockError = err;
+      if (attempt === OPEN_LOCK_RETRY_ATTEMPTS) break;
+      await sleep(OPEN_LOCK_RETRY_DELAY_MS * attempt);
+    }
+  }
+
+  // Final defense: only for recognized test fixtures, sweep stale sidecars
+  // (a prior aborted test run can leave a `.wal` lock that survives the
+  // tmp dir cleanup). Production paths never reach this branch — the guard
+  // requires the immediate parent dir to match a test prefix AND the
+  // resolved path to live under the OS temp directory.
+  if (isTestFixturePath(dbPath)) {
+    await sweepStaleSidecars(dbPath);
+    try {
+      return construct();
+    } catch {
+      // Intentionally do NOT overwrite originalLockError. The user-actionable
+      // signal is "we exhausted lock retries" — a different error from the
+      // post-sweep attempt is less useful than the lock failure that drove
+      // the sweep in the first place.
+    }
+  }
+  throw tagOpenRetryExhausted(originalLockError);
+};
+
 export async function openLbugConnection(
   lbugModule: LbugModule,
   databasePath: string,
@@ -74,7 +270,10 @@ export async function openLbugConnection(
 ): Promise<LbugConnectionHandle> {
   let db: lbug.Database | undefined;
   try {
-    db = createLbugDatabase(lbugModule, databasePath, options);
+    db = await openWithLockRetry(
+      () => createLbugDatabase(lbugModule, databasePath, options),
+      databasePath,
+    );
     return { db, conn: new lbugModule.Connection(db) };
   } catch (err) {
     if (db) await db.close().catch(() => {});
@@ -86,3 +285,60 @@ export async function closeLbugConnection(handle: LbugConnectionHandle): Promise
   await handle.conn.close().catch(() => {});
   await handle.db.close().catch(() => {});
 }
+
+/**
+ * Probe `dbPath` AND its `.wal` sidecar after `db.close()` so any
+ * residual native file handle surfaces as EBUSY/EPERM/EACCES and the
+ * bounded retry absorbs the release lag. Windows-only — Linux/macOS do
+ * not exhibit this race.
+ *
+ * Both files matter. Empirically, on rapid open→close→reopen cycles the
+ * main `dbPath` handle releases first; the `.wal` handle from the
+ * previous Database lingers and the new Database's first write (CREATE
+ * NODE TABLE during schema init) fails with "Could not set lock on
+ * file". Probing both makes safeClose actually return when the kernel
+ * is fully done with the path.
+ *
+ * Returns `true` when both probes succeeded (or skipped on non-lock
+ * errors / missing files). Returns `false` when either probe exhausted
+ * its budget with a lock code still in flight.
+ *
+ * Defensive shape:
+ *   - Opens read+write (`'r+'`) so the probe actually surfaces exclusive
+ *     locks held by the previous Database. A read-only probe (`'r'`) is
+ *     insufficient — Windows will grant read access while the previous
+ *     handle's exclusive write lock is still in flight, which lets
+ *     `safeClose` return before the next CREATE NODE TABLE can lock the
+ *     file.
+ *   - `try/finally` around `handle.close()` guarantees no fd leak even
+ *     if close itself throws.
+ */
+export const waitForWindowsHandleRelease = async (dbPath: string): Promise<boolean> => {
+  const mainReleased = await probeSinglePath(dbPath);
+  const walReleased = await probeSinglePath(dbPath + '.wal');
+  return mainReleased && walReleased;
+};
+
+const probeSinglePath = async (filePath: string): Promise<boolean> => {
+  for (let attempt = 1; attempt <= HANDLE_RELEASE_PROBE_ATTEMPTS; attempt++) {
+    let handle: fs.FileHandle | undefined;
+    try {
+      handle = await fs.open(filePath, 'r+');
+      return true;
+    } catch (err) {
+      const code = (err as NodeJS.ErrnoException | undefined)?.code;
+      if (!code || !HANDLE_RELEASE_LOCK_CODES.has(code)) return true; // ENOENT / unrelated → not our problem
+      if (attempt === HANDLE_RELEASE_PROBE_ATTEMPTS) return false;
+      await sleep(HANDLE_RELEASE_PROBE_DELAY_MS * attempt);
+    } finally {
+      if (handle) {
+        try {
+          await handle.close();
+        } catch {
+          /* swallow — caller cannot do anything useful with a probe-close failure */
+        }
+      }
+    }
+  }
+  return false;
+};
diff --git a/gitnexus/src/core/lbug/pool-adapter.ts b/gitnexus/src/core/lbug/pool-adapter.ts
index ca1c45611..ed999907e 100644
--- a/gitnexus/src/core/lbug/pool-adapter.ts
+++ b/gitnexus/src/core/lbug/pool-adapter.ts
@@ -18,7 +18,7 @@
 import fs from 'fs/promises';
 import lbug from '@ladybugdb/core';
 import { loadFTSExtension } from './lbug-adapter.js';
-import { createLbugDatabase } from './lbug-config.js';
+import { createLbugDatabase, isWalCorruptionError } from './lbug-config.js';
 
 /** Per-repo pool: one Database, many Connections */
 interface PoolEntry {
@@ -97,7 +97,7 @@ let idleTimer: ReturnType<typeof setInterval> | null = null;
 // @ladybugdb/core), corrupting stdout in the pre-sentinel window. Routing
 // through the leaf breaks that chain.
 export { realStdoutWrite, realStderrWrite, setActiveStdoutWrite } from '../../mcp/stdio-capture.js';
-import { getActiveStdoutWrite } from '../../mcp/stdio-capture.js';
+import { getActiveStdoutWrite, realStderrWrite } from '../../mcp/stdio-capture.js';
 
 let stdoutSilenceCount = 0;
 /** True while pre-warming connections — prevents watchdog from prematurely restoring stdout */
@@ -263,6 +263,46 @@ const WAITER_TIMEOUT_MS = 15_000;
 const LOCK_RETRY_ATTEMPTS = 3;
 const LOCK_RETRY_DELAY_MS = 2000;
 
+async function openReadOnlyDatabase(dbPath: string): Promise<lbug.Database> {
+  let db: lbug.Database | undefined;
+  silenceStdout();
+  try {
+    db = createLbugDatabase(lbug, dbPath, {
+      readOnly: true,
+      throwOnWalReplayFailure: false,
+    });
+    await db.init();
+    return db;
+  } catch (err) {
+    if (db) await db.close().catch(() => {});
+    throw err;
+  } finally {
+    restoreStdout();
+  }
+}
+
+/**
+ * Quarantine the .wal file and retry opening the database.
+ * Used when the initial open fails with a WAL corruption error.
+ */
+async function tryQuarantineAndReopen(dbPath: string, repoId: string): Promise<lbug.Database> {
+  const walPath = dbPath + '.wal';
+  const quarantineName = `${walPath}.corrupt.${Date.now()}-${Math.random().toString(36).slice(2)}`;
+  try {
+    await fs.rename(walPath, quarantineName);
+  } catch {
+    throw new Error(
+      `LadybugDB WAL corruption detected for ${repoId}. ` +
+        `Run \`gitnexus analyze\` to rebuild the index. (quarantine failed)`,
+    );
+  }
+  realStderrWrite(
+    `GitNexus: LadybugDB WAL quarantined for ${repoId}; graph may be stale. ` +
+      `Run \`gitnexus analyze\` to rebuild the index.\n`,
+  );
+  return await openReadOnlyDatabase(dbPath);
+}
+
 /** Deduplicates concurrent initLbug calls for the same repoId */
 const initPromises = new Map<string, Promise<void>>();
 
@@ -319,16 +359,29 @@ async function doInitLbug(repoId: string, dbPath: string): Promise<void> {
     // avoids lock conflicts when `gitnexus analyze` is writing.
     let lastError: Error | null = null;
     for (let attempt = 1; attempt <= LOCK_RETRY_ATTEMPTS; attempt++) {
-      silenceStdout();
       try {
-        const db = createLbugDatabase(lbug, dbPath, { readOnly: true });
-        restoreStdout();
+        const db = await openReadOnlyDatabase(dbPath);
         shared = { db, refCount: 0, ftsLoaded: false };
         dbCache.set(dbPath, shared);
         break;
       } catch (err: any) {
-        restoreStdout();
         lastError = err instanceof Error ? err : new Error(String(err));
+
+        if (isWalCorruptionError(lastError)) {
+          try {
+            const db = await tryQuarantineAndReopen(dbPath, repoId);
+            shared = { db, refCount: 0, ftsLoaded: false };
+            dbCache.set(dbPath, shared);
+            break;
+          } catch (retryErr) {
+            throw new Error(
+              `LadybugDB WAL corruption detected for ${repoId}. ` +
+                `Run \`gitnexus analyze\` to rebuild the index. ` +
+                `(${retryErr instanceof Error ? retryErr.message : String(retryErr)})`,
+            );
+          }
+        }
+
         const isLockError =
           lastError.message.includes('Could not set lock') || lastError.message.includes('lock');
         if (!isLockError || attempt === LOCK_RETRY_ATTEMPTS) break;
diff --git a/gitnexus/src/core/lbug/schema.ts b/gitnexus/src/core/lbug/schema.ts
index 5579036f1..cd6838dd8 100644
--- a/gitnexus/src/core/lbug/schema.ts
+++ b/gitnexus/src/core/lbug/schema.ts
@@ -167,7 +167,18 @@ export const TYPE_ALIAS_SCHEMA = CODE_ELEMENT_BASE('TypeAlias');
 export const CONST_SCHEMA = CODE_ELEMENT_BASE('Const');
 export const STATIC_SCHEMA = CODE_ELEMENT_BASE('Static');
 export const VARIABLE_SCHEMA = CODE_ELEMENT_BASE('Variable');
-export const PROPERTY_SCHEMA = CODE_ELEMENT_BASE('Property');
+export const PROPERTY_SCHEMA = `
+CREATE NODE TABLE \`Property\` (
+  id STRING,
+  name STRING,
+  filePath STRING,
+  startLine INT64,
+  endLine INT64,
+  content STRING,
+  description STRING,
+  declaredType STRING,
+  PRIMARY KEY (id)
+)`;
 export const RECORD_SCHEMA = CODE_ELEMENT_BASE('Record');
 export const DELEGATE_SCHEMA = CODE_ELEMENT_BASE('Delegate');
 export const ANNOTATION_SCHEMA = CODE_ELEMENT_BASE('Annotation');
diff --git a/gitnexus/src/core/search/bm25-index.ts b/gitnexus/src/core/search/bm25-index.ts
index 58c0343c9..27a7b9d8d 100644
--- a/gitnexus/src/core/search/bm25-index.ts
+++ b/gitnexus/src/core/search/bm25-index.ts
@@ -15,9 +15,16 @@ export interface BM25SearchResult {
   nodeIds?: string[];
 }
 
+export interface FTSSearchResponse {
+  results: BM25SearchResult[];
+  /** True when at least one FTS index query succeeded (index exists). */
+  ftsAvailable: boolean;
+}
+
 /**
  * Execute a single FTS query via a custom executor (for MCP connection pool).
- * Returns the same shape as core queryFTS (from LadybugDB adapter).
+ * Returns `null` when the query fails (e.g. FTS index does not exist) so the
+ * caller can distinguish "zero matches" from "index missing".
  */
 async function queryFTSViaExecutor(
   executor: (cypher: string) => Promise<any[]>,
@@ -25,7 +32,7 @@ async function queryFTSViaExecutor(
   indexName: string,
   query: string,
   limit: number,
-): Promise<Array<{ filePath: string; score: number; nodeId: string }>> {
+): Promise<Array<{ filePath: string; score: number; nodeId: string }> | null> {
   // Escape single quotes and backslashes to prevent Cypher injection
   const escapedQuery = query.replace(/\\/g, '\\\\').replace(/'/g, "''");
   const cypher = `
@@ -46,7 +53,7 @@ async function queryFTSViaExecutor(
       };
     });
   } catch {
-    return [];
+    return null;
   }
 }
 
@@ -65,8 +72,9 @@ export const searchFTSFromLbug = async (
   query: string,
   limit: number = 20,
   repoId?: string,
-): Promise<BM25SearchResult[]> => {
+): Promise<FTSSearchResponse> => {
   const resultsByIndex: any[][] = [];
+  let queriesSucceeded = 0;
 
   if (repoId) {
     // Use MCP connection pool via dynamic import
@@ -77,15 +85,27 @@ export const searchFTSFromLbug = async (
     const executor = (cypher: string) => executeQuery(repoId, cypher);
 
     for (const { table, indexName } of FTS_INDEXES) {
-      resultsByIndex.push(await queryFTSViaExecutor(executor, table, indexName, query, limit));
+      const result = await queryFTSViaExecutor(executor, table, indexName, query, limit);
+      if (result !== null) {
+        queriesSucceeded++;
+        resultsByIndex.push(result);
+      }
     }
   } else {
     // Use core lbug adapter (CLI / pipeline context) — also sequential for safety.
     for (const { table, indexName } of FTS_INDEXES) {
-      resultsByIndex.push(await queryFTS(table, indexName, query, limit, false).catch(() => []));
+      try {
+        const result = await queryFTS(table, indexName, query, limit, false);
+        queriesSucceeded++;
+        resultsByIndex.push(result);
+      } catch {
+        // FTS index may not exist — count as failed
+      }
     }
   }
 
+  const ftsAvailable = queriesSucceeded > 0;
+
   // Collect all node scores per filePath to track which nodes actually matched
   const fileNodeScores = new Map<string, Array<{ score: number; nodeId: string }>>();
 
@@ -116,10 +136,13 @@ export const searchFTSFromLbug = async (
     .sort((a, b) => b.score - a.score)
     .slice(0, limit);
 
-  return sorted.map((r, index) => ({
-    filePath: r.filePath,
-    score: r.score,
-    rank: index + 1,
-    nodeIds: r.nodeIds,
-  }));
+  return {
+    results: sorted.map((r, index) => ({
+      filePath: r.filePath,
+      score: r.score,
+      rank: index + 1,
+      nodeIds: r.nodeIds,
+    })),
+    ftsAvailable,
+  };
 };
diff --git a/gitnexus/src/core/search/hybrid-search.ts b/gitnexus/src/core/search/hybrid-search.ts
index 72dd1c9b5..b76a9f5e9 100644
--- a/gitnexus/src/core/search/hybrid-search.ts
+++ b/gitnexus/src/core/search/hybrid-search.ts
@@ -113,12 +113,13 @@ export const mergeWithRRF = (
 };
 
 /**
- * Check if hybrid search is available
- * LadybugDB FTS is always available once the database is initialized.
- * Semantic search is optional - hybrid works with just FTS if embeddings aren't ready.
+ * Check if hybrid search is available.
+ * FTS indexes may be missing on read-only MCP connections (see #1403);
+ * callers should inspect `ftsAvailable` from searchFTSFromLbug for
+ * per-query availability. This helper is a coarse gate only.
  */
 export const isHybridSearchReady = (): boolean => {
-  return true; // FTS is always available via LadybugDB when DB is open
+  return true; // FTS is attempted on every query; ftsAvailable signals actual availability
 };
 
 /**
@@ -160,7 +161,7 @@ export const hybridSearch = async (
   ) => Promise<SemanticSearchResult[]>,
 ): Promise<HybridSearchResult[]> => {
   // Use LadybugDB FTS for always-fresh BM25 results
-  const bm25Results = await searchFTSFromLbug(query, limit);
+  const { results: bm25Results } = await searchFTSFromLbug(query, limit);
   const semanticResults = await semanticSearch(executeQuery, query, limit);
   return mergeWithRRF(bm25Results, semanticResults, limit);
 };
diff --git a/gitnexus/src/mcp/core/embedder.ts b/gitnexus/src/mcp/core/embedder.ts
index ff3a12fd8..f529f2805 100644
--- a/gitnexus/src/mcp/core/embedder.ts
+++ b/gitnexus/src/mcp/core/embedder.ts
@@ -12,7 +12,11 @@ import {
   httpEmbedQuery,
 } from '../../core/embeddings/http-client.js';
 import { resolveEmbeddingConfig } from '../../core/embeddings/config.js';
-import { applyHfEnvOverrides } from '../../core/embeddings/hf-env.js';
+import {
+  applyHfEnvOverrides,
+  isHfDownloadFailure,
+  withHfDownloadRetry,
+} from '../../core/embeddings/hf-env.js';
 import { silenceStdout, restoreStdout, realStderrWrite } from '../../core/lbug/pool-adapter.js';
 
 import { logger } from '../../core/logger.js';
@@ -69,23 +73,39 @@ export const initEmbedder = async (): Promise<FeatureExtractionPipeline> => {
           silenceStdout();
           process.stderr.write = (() => true) as any;
           try {
-            embedderInstance = await (pipeline as any)('feature-extraction', MODEL_ID, {
-              device: device,
-              dtype: 'fp32',
-              session_options: {
-                logSeverityLevel: 3,
-                intraOpNumThreads: embeddingConfig.threads,
-                interOpNumThreads: 1,
-                executionMode: 'sequential',
-              },
-            });
+            embedderInstance = await withHfDownloadRetry(() =>
+              pipeline('feature-extraction', MODEL_ID, {
+                device: device,
+                dtype: 'fp32',
+                session_options: {
+                  logSeverityLevel: 3,
+                  intraOpNumThreads: embeddingConfig.threads,
+                  interOpNumThreads: 1,
+                  executionMode: 'sequential',
+                },
+              }),
+            );
           } finally {
             restoreStdout();
             process.stderr.write = realStderrWrite;
           }
           logger.info({ device }, 'GitNexus: Embedding model loaded');
           return embedderInstance!;
-        } catch {
+        } catch (deviceError) {
+          // Network errors and circuit-open errors are not device-specific —
+          // they will fail the same way on every device. Rethrow immediately
+          // with actionable HF_ENDPOINT guidance rather than silently falling
+          // back to the next device.
+          const errMsg = deviceError instanceof Error ? deviceError.message : String(deviceError);
+          if (isHfDownloadFailure(errMsg)) {
+            const endpointHint = process.env.HF_ENDPOINT
+              ? `The configured endpoint (${process.env.HF_ENDPOINT}) may be unreachable.`
+              : `huggingface.co may be unreachable from your network.\n` +
+                `  Set HF_ENDPOINT to a mirror and retry:\n` +
+                `    HF_ENDPOINT=https://hf-mirror.com npx gitnexus analyze --embeddings\n` +
+                `    (Windows: set HF_ENDPOINT=https://hf-mirror.com && npx gitnexus analyze --embeddings)`;
+            throw new Error(`Failed to download embedding model: ${errMsg}\n  ${endpointHint}`);
+          }
           if (device === 'cpu') throw new Error('Failed to load embedding model');
         }
       }
diff --git a/gitnexus/src/mcp/local/local-backend.ts b/gitnexus/src/mcp/local/local-backend.ts
index 11f7e61f3..167c1db81 100644
--- a/gitnexus/src/mcp/local/local-backend.ts
+++ b/gitnexus/src/mcp/local/local-backend.ts
@@ -16,6 +16,7 @@ import {
   isLbugReady,
   isWriteQuery,
 } from '../../core/lbug/pool-adapter.js';
+import { isWalCorruptionError, WAL_RECOVERY_SUGGESTION } from '../../core/lbug/lbug-config.js';
 export { isWriteQuery };
 // Embedding imports are lazy (dynamic import) to avoid loading onnxruntime-node
 // at MCP server startup — crashes on unsupported Node ABI versions (#89)
@@ -40,7 +41,7 @@ import {
   isVectorExtensionSupportedByPlatform,
 } from '../../core/platform/capabilities.js';
 import { PhaseTimer } from '../../core/search/phase-timer.js';
-import { checkStaleness, checkCwdMatch } from '../../core/git-staleness.js';
+import { checkStalenessAsync, checkCwdMatch } from '../../core/git-staleness.js';
 import { logger } from '../../core/logger.js';
 // AI context generation is CLI-only (gitnexus analyze)
 // import { generateAIContextFiles } from '../../cli/ai-context.js';
@@ -554,8 +555,15 @@ export class LocalBackend {
       byRemote.set(h.remoteUrl, list);
     }
 
-    return handles.map((h) => {
-      const stale = checkStaleness(h.repoPath, h.lastCommit);
+    // Check staleness for all repos in parallel instead of sequentially.
+    // Each check spawns an async `git rev-list` — with 200 repos the sync
+    // variant took ~50 s; parallel async brings it under a second (#1363).
+    const stalenessResults = await Promise.all(
+      handles.map((h) => checkStalenessAsync(h.repoPath, h.lastCommit)),
+    );
+
+    return handles.map((h, i) => {
+      const stale = stalenessResults[i];
       const selfNorm = norm(h.repoPath);
       const siblings = h.remoteUrl
         ? (byRemote.get(h.remoteUrl) ?? []).filter((e) => norm(e.repoPath) !== selfNorm)
@@ -971,7 +979,7 @@ export class LocalBackend {
       timing,
       ...(!ftsUsed && {
         warning:
-          'FTS extension unavailable - keyword search degraded. Run: gitnexus analyze --force to rebuild indexes.',
+          'FTS indexes missing — keyword search degraded. Run: gitnexus analyze --force to rebuild indexes.',
       }),
     };
   }
@@ -985,9 +993,9 @@ export class LocalBackend {
     limit: number,
   ): Promise<{ results: any[]; ftsUsed: boolean }> {
     const { searchFTSFromLbug } = await import('../../core/search/bm25-index.js');
-    let bm25Results;
+    let ftsResponse;
     try {
-      bm25Results = await searchFTSFromLbug(query, limit, repo.id);
+      ftsResponse = await searchFTSFromLbug(query, limit, repo.id);
     } catch (err: any) {
       logger.error(
         { err: err.message },
@@ -996,7 +1004,8 @@ export class LocalBackend {
       return { results: [], ftsUsed: false };
     }
 
-    const ftsUsed = bm25Results.length === 0 || bm25Results[0]?.ftsUsed !== false;
+    const bm25Results = ftsResponse.results;
+    const ftsUsed = ftsResponse.ftsAvailable;
 
     const results: any[] = [];
 
@@ -1218,7 +1227,14 @@ export class LocalBackend {
       const result = await executeQuery(repo.id, params.query);
       return result;
     } catch (err: any) {
-      return { error: err.message || 'Query failed' };
+      const msg = err.message || 'Query failed';
+      if (isWalCorruptionError(err)) {
+        return {
+          error: msg,
+          recoverySuggestion: WAL_RECOVERY_SUGGESTION,
+        };
+      }
+      return { error: msg };
     }
   }
 
@@ -1672,6 +1688,30 @@ export class LocalBackend {
       kind?: string;
       include_content?: boolean;
     },
+  ): Promise<any> {
+    try {
+      return await this._contextImpl(repo, params);
+    } catch (err: any) {
+      const msg = (err instanceof Error ? err.message : String(err)) || 'Context query failed';
+      if (isWalCorruptionError(err)) {
+        return {
+          error: msg,
+          recoverySuggestion: WAL_RECOVERY_SUGGESTION,
+        };
+      }
+      throw err;
+    }
+  }
+
+  private async _contextImpl(
+    repo: RepoHandle,
+    params: {
+      name?: string;
+      uid?: string;
+      file_path?: string;
+      kind?: string;
+      include_content?: boolean;
+    },
   ): Promise<any> {
     await this.ensureInitialized(repo.id);
 
@@ -1716,12 +1756,13 @@ export class LocalBackend {
       repo.id,
       `
       MATCH (caller)-[r:CodeRelation]->(n {id: $symId})
-      WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'HAS_METHOD', 'HAS_PROPERTY', 'METHOD_OVERRIDES', 'OVERRIDES', 'METHOD_IMPLEMENTS', 'ACCESSES']
+      WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'USES', 'HAS_METHOD', 'HAS_PROPERTY', 'METHOD_OVERRIDES', 'OVERRIDES', 'METHOD_IMPLEMENTS', 'ACCESSES']
       RETURN r.type AS relType, caller.id AS uid, caller.name AS name, caller.filePath AS filePath, labels(caller)[0] AS kind
       LIMIT 30
     `,
       { symId },
     );
+    let typedPropertyRows: any[] = [];
 
     // Fix #480: Class/Interface nodes have no direct CALLS/IMPORTS edges —
     // those point to Constructor and File nodes respectively. Fetch those
@@ -1755,23 +1796,24 @@ export class LocalBackend {
 
     if (isClassLike) {
       try {
-        // Run both incoming-ref queries in parallel — they are independent.
-        const [ctorIncoming, fileIncoming] = await Promise.all([
-          executeParameterized(
-            repo.id,
-            `
+        // Run incoming-ref queries in parallel — they are independent.
+        const [ctorIncoming, fileIncoming, typedPropertyIncoming, typedProperties] =
+          await Promise.all([
+            executeParameterized(
+              repo.id,
+              `
             MATCH (n)-[hm:CodeRelation]->(ctor:Constructor)
             WHERE n.id = $symId AND hm.type = 'HAS_METHOD'
             MATCH (caller)-[r:CodeRelation]->(ctor)
-            WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'ACCESSES']
+            WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'USES', 'ACCESSES']
             RETURN r.type AS relType, caller.id AS uid, caller.name AS name, caller.filePath AS filePath, labels(caller)[0] AS kind
             LIMIT 30
           `,
-            { symId },
-          ),
-          executeParameterized(
-            repo.id,
-            `
+              { symId },
+            ),
+            executeParameterized(
+              repo.id,
+              `
             MATCH (f:File)-[rel:CodeRelation]->(n)
             WHERE n.id = $symId AND rel.type = 'DEFINES'
             MATCH (caller)-[r:CodeRelation]->(f)
@@ -1779,9 +1821,45 @@ export class LocalBackend {
             RETURN r.type AS relType, caller.id AS uid, caller.name AS name, caller.filePath AS filePath, labels(caller)[0] AS kind
             LIMIT 30
           `,
-            { symId },
-          ),
-        ]);
+              { symId },
+            ),
+            executeParameterized(
+              repo.id,
+              `
+            MATCH (p:\`Property\`)
+            WHERE p.declaredType = $name
+               OR p.declaredType STARTS WITH $genericPrefix
+               OR p.declaredType CONTAINS $genericArg
+            MATCH (caller)-[r:CodeRelation]->(p)
+            WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'USES', 'ACCESSES']
+            RETURN r.type AS relType, caller.id AS uid, caller.name AS name, caller.filePath AS filePath, labels(caller)[0] AS kind
+            LIMIT 30
+          `,
+              {
+                name: sym.name,
+                genericPrefix: `${sym.name}<`,
+                genericArg: `<${sym.name}>`,
+              },
+            ),
+            executeParameterized(
+              repo.id,
+              `
+            MATCH (p:\`Property\`)
+            WHERE p.declaredType = $name
+               OR p.declaredType STARTS WITH $genericPrefix
+               OR p.declaredType CONTAINS $genericArg
+            RETURN p.id AS uid, p.name AS name, p.filePath AS filePath, labels(p)[0] AS kind,
+                   p.declaredType AS declaredType
+            LIMIT 30
+          `,
+              {
+                name: sym.name,
+                genericPrefix: `${sym.name}<`,
+                genericArg: `<${sym.name}>`,
+              },
+            ),
+          ]);
+        typedPropertyRows = typedProperties;
 
         // Deduplicate by (relType, uid) — a caller can have multiple relation
         // types to the same target (e.g. both IMPORTS and CALLS), and each
@@ -1789,7 +1867,7 @@ export class LocalBackend {
         const seenKeys = new Set(
           incomingRows.map((r: any) => `${r.relType || r[0]}:${r.uid || r[1]}`),
         );
-        for (const r of [...ctorIncoming, ...fileIncoming]) {
+        for (const r of [...ctorIncoming, ...fileIncoming, ...typedPropertyIncoming]) {
           const key = `${r.relType || r[0]}:${r.uid || r[1]}`;
           if (!seenKeys.has(key)) {
             seenKeys.add(key);
@@ -1806,7 +1884,7 @@ export class LocalBackend {
       repo.id,
       `
       MATCH (n {id: $symId})-[r:CodeRelation]->(target)
-      WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'HAS_METHOD', 'HAS_PROPERTY', 'METHOD_OVERRIDES', 'OVERRIDES', 'METHOD_IMPLEMENTS', 'ACCESSES']
+      WHERE r.type IN ['CALLS', 'IMPORTS', 'EXTENDS', 'IMPLEMENTS', 'USES', 'HAS_METHOD', 'HAS_PROPERTY', 'METHOD_OVERRIDES', 'OVERRIDES', 'METHOD_IMPLEMENTS', 'ACCESSES']
       RETURN r.type AS relType, target.id AS uid, target.name AS name, target.filePath AS filePath, labels(target)[0] AS kind
       LIMIT 30
     `,
@@ -1895,6 +1973,17 @@ export class LocalBackend {
       },
       incoming: categorize(incomingRows),
       outgoing: categorize(outgoingRows),
+      ...(typedPropertyRows.length > 0
+        ? {
+            typed_properties: typedPropertyRows.map((r: any) => ({
+              uid: r.uid || r[0],
+              name: r.name || r[1],
+              filePath: r.filePath || r[2],
+              kind: r.kind || r[3],
+              declaredType: r.declaredType || r[4],
+            })),
+          }
+        : {}),
       processes: processRows.map((r: any) => ({
         id: r.pid || r[0],
         name: r.label || r[1],
@@ -2433,6 +2522,7 @@ export class LocalBackend {
         impactedCount: 0,
         risk: 'UNKNOWN',
         suggestion: 'The graph query failed — try gitnexus context <symbol> as a fallback',
+        ...(isWalCorruptionError(err) ? { recoverySuggestion: WAL_RECOVERY_SUGGESTION } : {}),
       };
     }
   }
@@ -2459,6 +2549,7 @@ export class LocalBackend {
     const mappedRelTypes = params.relationTypes?.flatMap((t: string) =>
       t === 'OVERRIDES' ? ['OVERRIDES', 'METHOD_OVERRIDES'] : [t],
     );
+    const hasExplicitRelationTypes = mappedRelTypes !== undefined && mappedRelTypes.length > 0;
     const rawRelTypes =
       mappedRelTypes && mappedRelTypes.length > 0
         ? mappedRelTypes.filter((t: string) => VALID_RELATION_TYPES.has(t))
@@ -2467,6 +2558,7 @@ export class LocalBackend {
             'IMPORTS',
             'EXTENDS',
             'IMPLEMENTS',
+            'USES',
             'METHOD_OVERRIDES',
             'OVERRIDES',
             'METHOD_IMPLEMENTS',
@@ -2479,6 +2571,7 @@ export class LocalBackend {
             'IMPORTS',
             'EXTENDS',
             'IMPLEMENTS',
+            'USES',
             'METHOD_OVERRIDES',
             'OVERRIDES',
             'METHOD_IMPLEMENTS',
@@ -2538,9 +2631,16 @@ export class LocalBackend {
     };
     const symType = outcome.resolvedLabel || outcome.symbol.type || '';
 
+    const effectiveRelationTypes =
+      (symType === 'Class' || symType === 'Interface') &&
+      !hasExplicitRelationTypes &&
+      !relationTypes.includes('ACCESSES')
+        ? [...relationTypes, 'ACCESSES']
+        : relationTypes;
+
     return this._runImpactBFS(repo, sym, symType, direction, {
       maxDepth,
-      relationTypes,
+      relationTypes: effectiveRelationTypes,
       includeTests,
       minConfidence,
     });
@@ -2619,6 +2719,30 @@ export class LocalBackend {
             frontier.push(rid);
           }
         }
+
+        const typedPropertyRows = await executeParameterized(
+          repo.id,
+          `
+          MATCH (p:\`Property\`)
+          WHERE p.declaredType = $name
+             OR p.declaredType STARTS WITH $genericPrefix
+             OR p.declaredType CONTAINS $genericArg
+          RETURN p.id AS id, p.name AS name, labels(p)[0] AS type, p.filePath AS filePath
+        `,
+          {
+            name: sym.name,
+            genericPrefix: `${sym.name}<`,
+            genericArg: `<${sym.name}>`,
+          },
+        );
+
+        for (const r of typedPropertyRows) {
+          const rid = r.id || r[0];
+          if (rid && !visited.has(rid)) {
+            visited.add(rid);
+            frontier.push(rid);
+          }
+        }
       } catch (e) {
         logQueryError('impact:class-node-expansion', e);
       }
@@ -2984,8 +3108,14 @@ export class LocalBackend {
       relationTypes: string[];
       minConfidence: number;
       includeTests: boolean;
+      signal?: AbortSignal;
     },
   ): Promise<any | null> {
+    // Honor an already-aborted signal at the entry boundary as a fast
+    // path. Cooperative cancellation inside _runImpactBFS is out of
+    // scope — the caller's Promise.race against the same signal
+    // resolves the await regardless of how long this body runs.
+    if (opts.signal?.aborted) return null;
     try {
       await this.refreshRepos();
       await this.ensureInitialized(repoId);
diff --git a/gitnexus/src/server/api.ts b/gitnexus/src/server/api.ts
index 773190785..cc65daa3d 100644
--- a/gitnexus/src/server/api.ts
+++ b/gitnexus/src/server/api.ts
@@ -1060,11 +1060,12 @@ export const createServer = async (port: number, host: string = '127.0.0.1') =>
 
       const results = await withLbugDb(lbugPath, async () => {
         let searchResults: any[];
+        let ftsAvailable: boolean | undefined;
 
         if (mode === 'semantic') {
           const { isEmbedderReady } = await import('../core/embeddings/embedder.js');
           if (!isEmbedderReady()) {
-            return [] as any[];
+            return { searchResults: [] as any[], ftsAvailable: undefined };
           }
           const { semanticSearch: semSearch } =
             await import('../core/embeddings/embedding-pipeline.js');
@@ -1077,8 +1078,9 @@ export const createServer = async (port: number, host: string = '127.0.0.1') =>
             sources: ['semantic'],
           }));
         } else if (mode === 'bm25') {
-          searchResults = await searchFTSFromLbug(query, limit);
-          searchResults = searchResults.map((r: any, i: number) => ({
+          const ftsResponse = await searchFTSFromLbug(query, limit);
+          ftsAvailable = ftsResponse.ftsAvailable;
+          searchResults = ftsResponse.results.map((r: any, i: number) => ({
             ...r,
             rank: i + 1,
             sources: ['bm25'],
@@ -1091,11 +1093,13 @@ export const createServer = async (port: number, host: string = '127.0.0.1') =>
               await import('../core/embeddings/embedding-pipeline.js');
             searchResults = await hybridSearch(query, limit, executeQuery, semSearch);
           } else {
-            searchResults = await searchFTSFromLbug(query, limit);
+            const ftsResponse = await searchFTSFromLbug(query, limit);
+            ftsAvailable = ftsResponse.ftsAvailable;
+            searchResults = ftsResponse.results;
           }
         }
 
-        if (!enrich) return searchResults;
+        if (!enrich) return { searchResults, ftsAvailable };
 
         // Server-side enrichment: add connections, cluster, processes per result
         // Uses parameterized queries to prevent Cypher injection via nodeId
@@ -1177,9 +1181,14 @@ export const createServer = async (port: number, host: string = '127.0.0.1') =>
           }),
         );
 
-        return enriched;
+        return { searchResults: enriched, ftsAvailable };
       });
-      res.json({ results });
+      const response: any = { results: results.searchResults ?? results };
+      if (results.ftsAvailable === false) {
+        response.warning =
+          'FTS indexes missing — keyword search degraded. Run: gitnexus analyze --force to rebuild indexes.';
+      }
+      res.json(response);
     } catch (err: any) {
       res.status(500).json({ error: err.message || 'Search failed' });
     }
diff --git a/gitnexus/test/fixtures/lang-resolution/csharp-generic-type-refs/Program.cs b/gitnexus/test/fixtures/lang-resolution/csharp-generic-type-refs/Program.cs
new file mode 100644
index 000000000..28674ecda
--- /dev/null
+++ b/gitnexus/test/fixtures/lang-resolution/csharp-generic-type-refs/Program.cs
@@ -0,0 +1,18 @@
+namespace App;
+
+public class USER_INFO
+{
+    public string? USER_ID { get; set; }
+}
+
+public interface IEntityTypeConfiguration<T>
+{
+}
+
+public class UserInfoConfiguration : IEntityTypeConfiguration<USER_INFO>
+{
+    public Task<List<USER_INFO>> Load(List<USER_INFO> users)
+    {
+        return Task.FromResult(users);
+    }
+}
diff --git a/gitnexus/test/helpers/test-db.ts b/gitnexus/test/helpers/test-db.ts
index 5818fdc8e..37032ebc8 100644
--- a/gitnexus/test/helpers/test-db.ts
+++ b/gitnexus/test/helpers/test-db.ts
@@ -37,6 +37,13 @@ export async function cleanupTempDir(tmpDir: string): Promise<void> {
 /**
  * Create a temporary directory for LadybugDB tests.
  * Returns the path and a cleanup function.
+ *
+ * IMPORTANT: when adding a new test that passes a custom `prefix`, also add
+ * the prefix to `TEST_FIXTURE_PREFIXES` in
+ * `gitnexus/src/core/lbug/lbug-config.ts`. The stale-sidecar sweep relies
+ * on the prefix list to recognize test fixtures; an unknown prefix means
+ * the sweep silently won't fire for that fixture and Windows CI flakes
+ * return.
  */
 export async function createTempDir(prefix: string = 'gitnexus-test-'): Promise<TestDBHandle> {
   const tmpDir = await fs.mkdtemp(path.join(os.tmpdir(), prefix));
diff --git a/gitnexus/test/integration/context-typed-property.test.ts b/gitnexus/test/integration/context-typed-property.test.ts
new file mode 100644
index 000000000..aef442147
--- /dev/null
+++ b/gitnexus/test/integration/context-typed-property.test.ts
@@ -0,0 +1,77 @@
+/**
+ * Integration test: context() expands Class symbols through typed properties.
+ *
+ * Reproduces EF-style usage where code reads a DbContext property
+ * (`db.USER_INFO`) whose source type is `DbSet<USER_INFO>`. The direct
+ * graph edge is Method -> Property, not Method -> Class, so context() must
+ * use the same typed-property bridge that impact() uses.
+ */
+import { beforeAll, describe, expect, it, vi } from 'vitest';
+import { LocalBackend } from '../../src/mcp/local/local-backend.js';
+import { listRegisteredRepos } from '../../src/storage/repo-manager.js';
+import { withTestLbugDB } from '../helpers/test-indexed-db.js';
+
+vi.mock('../../src/storage/repo-manager.js', () => ({
+  listRegisteredRepos: vi.fn().mockResolvedValue([]),
+  cleanupOldKuzuFiles: vi.fn().mockResolvedValue({ found: false, needsReindex: false }),
+  findSiblingClones: vi.fn().mockResolvedValue([]),
+}));
+
+const SEED = [
+  `CREATE (c:Class {id:'Class:Models/USER_INFO.cs:USER_INFO', name:'USER_INFO', filePath:'Models/USER_INFO.cs', startLine:1, endLine:5, content:'public class USER_INFO {}', description:''})`,
+  `CREATE (p:\`Property\` {id:'Property:Data/UserDbContext.cs:UserDbContext.USER_INFO', name:'USER_INFO', filePath:'Data/UserDbContext.cs', startLine:10, endLine:10, content:'public DbSet<USER_INFO> USER_INFO { get; set; }', description:'', declaredType:'DbSet<USER_INFO>'})`,
+  `CREATE (m:Method {id:'Method:Services/UserService.cs:UserService.GetUserInfo#1', name:'GetUserInfo', filePath:'Services/UserService.cs', startLine:20, endLine:30, isExported:false, content:'db.USER_INFO.FirstOrDefault();', description:'', parameterCount:1, returnType:'USER_INFO'})`,
+  `MATCH (m:Method {id:'Method:Services/UserService.cs:UserService.GetUserInfo#1'}), (p:\`Property\` {id:'Property:Data/UserDbContext.cs:UserDbContext.USER_INFO'}) CREATE (m)-[:CodeRelation {type:'ACCESSES', confidence:1.0, reason:'read', step:1}]->(p)`,
+];
+
+withTestLbugDB(
+  'context-typed-property',
+  (handle) => {
+    let backend: LocalBackend;
+
+    beforeAll(async () => {
+      backend = (handle as any)._backend;
+    });
+
+    describe('context() typed-property expansion', () => {
+      it('surfaces property callers and explains the typed property bridge', async () => {
+        const result = await backend.callTool('context', {
+          uid: 'Class:Models/USER_INFO.cs:USER_INFO',
+        });
+
+        expect(result.status).toBe('found');
+        expect(result.symbol.kind).toBe('Class');
+
+        const accesses = result.incoming.accesses || [];
+        expect(accesses.map((r: any) => r.name)).toContain('GetUserInfo');
+
+        expect(result.typed_properties).toEqual([
+          expect.objectContaining({
+            uid: 'Property:Data/UserDbContext.cs:UserDbContext.USER_INFO',
+            name: 'USER_INFO',
+            declaredType: 'DbSet<USER_INFO>',
+          }),
+        ]);
+      });
+    });
+  },
+  {
+    seed: SEED,
+    poolAdapter: true,
+    afterSetup: async (handle) => {
+      vi.mocked(listRegisteredRepos).mockResolvedValue([
+        {
+          name: 'test-repo',
+          path: '/test/repo',
+          storagePath: handle.tmpHandle.dbPath,
+          indexedAt: new Date().toISOString(),
+          lastCommit: 'abc123',
+          stats: { files: 3, nodes: 3, communities: 0, processes: 0 },
+        },
+      ]);
+      const backend = new LocalBackend();
+      await backend.init();
+      (handle as any)._backend = backend;
+    },
+  },
+);
diff --git a/gitnexus/test/integration/group/include-extractor-sync.test.ts b/gitnexus/test/integration/group/include-extractor-sync.test.ts
new file mode 100644
index 000000000..908664bdc
--- /dev/null
+++ b/gitnexus/test/integration/group/include-extractor-sync.test.ts
@@ -0,0 +1,195 @@
+/**
+ * Integration test: IncludeExtractor output → group matching → bridge DB.
+ *
+ * Covers PR #1156 review finding #7: verifies that the full runtime path
+ * (IncludeExtractor → StoredContract → runExactMatch → CrossLinks → writeBridge)
+ * stays wired up. A regression in either normalizeContractId or the include
+ * branch of ManifestExtractor.resolveSymbol would produce 0 cross-links and
+ * fail this test.
+ */
+import { describe, it, expect } from 'vitest';
+import * as fs from 'node:fs';
+import * as os from 'node:os';
+import * as path from 'node:path';
+import { parseGroupConfig } from '../../../src/core/group/config-parser.js';
+import { syncGroup } from '../../../src/core/group/sync.js';
+import type { StoredContract } from '../../../src/core/group/types.js';
+import { IncludeExtractor } from '../../../src/core/group/extractors/include-extractor.js';
+import { normalizeContractId } from '../../../src/core/group/matching.js';
+
+const GROUP_YAML = [
+  'version: 1',
+  'name: include-test-group',
+  'description: "IncludeExtractor integration test"',
+  '',
+  'repos:',
+  '  app/provider: include-provider',
+  '  app/consumer: include-consumer',
+  '',
+  'links: []',
+  'packages: {}',
+  '',
+  'detect:',
+  '  http: false',
+  '  grpc: false',
+  '  topics: false',
+  '  shared_libs: false',
+  '  includes: true',
+  '  embedding_fallback: false',
+  '',
+  'matching:',
+  '  bm25_threshold: 0.7',
+  '  embedding_threshold: 0.65',
+  '  max_candidates_per_step: 3',
+].join('\n');
+
+describe('IncludeExtractor → syncGroup integration (finding #7)', () => {
+  it('produces a CrossLink when provider and consumer emit the same include contract-id', async () => {
+    const config = parseGroupConfig(GROUP_YAML);
+
+    // Mock the IncludeExtractor output directly — a header provider in one
+    // repo and a quoted #include consumer in the other, both normalized to
+    // the same include::map/base/view.h contract-id.
+    const mockContracts: StoredContract[] = [
+      {
+        contractId: 'include::map/base/view.h',
+        type: 'include',
+        role: 'provider',
+        symbolUid: 'File:map/base/view.h',
+        symbolRef: { filePath: 'map/base/view.h', name: 'view.h' },
+        symbolName: 'view.h',
+        confidence: 0.95,
+        meta: { source: 'filesystem' },
+        repo: 'app/provider',
+      },
+      {
+        contractId: 'include::map/base/view.h',
+        type: 'include',
+        role: 'consumer',
+        symbolUid: 'File:src/controller.cpp',
+        symbolRef: { filePath: 'src/controller.cpp', name: 'map/base/view.h' },
+        symbolName: 'map/base/view.h',
+        confidence: 0.85,
+        meta: { source: 'tree_sitter', includePath: 'map/base/view.h' },
+        repo: 'app/consumer',
+      },
+    ];
+
+    const result = await syncGroup(config, {
+      extractorOverride: async () => mockContracts,
+      skipWrite: true,
+    });
+
+    const includeLinks = result.crossLinks.filter((l) => l.type === 'include');
+    expect(includeLinks.length).toBeGreaterThanOrEqual(1);
+
+    const link = includeLinks[0];
+    expect(link.contractId).toBe('include::map/base/view.h');
+    expect(link.matchType).toBe('exact');
+    expect(link.from.repo).toBe('app/consumer');
+    expect(link.to.repo).toBe('app/provider');
+  });
+
+  it('normalizes mixed-case / backslash include paths to the same contract-id end-to-end', async () => {
+    const config = parseGroupConfig(GROUP_YAML);
+
+    // Provider writes the canonical form; consumer's include has mixed case
+    // and a backslash. After normalizeContractId they must still match.
+    const providerId = 'include::map/base/view.h';
+    const rawConsumerId = 'include::Map\\Base\\View.h';
+
+    // Sanity — normalizeContractId must collapse them.
+    expect(normalizeContractId(rawConsumerId)).toBe(providerId);
+
+    const mockContracts: StoredContract[] = [
+      {
+        contractId: providerId,
+        type: 'include',
+        role: 'provider',
+        symbolUid: 'File:map/base/view.h',
+        symbolRef: { filePath: 'map/base/view.h', name: 'view.h' },
+        symbolName: 'view.h',
+        confidence: 0.95,
+        meta: { source: 'filesystem' },
+        repo: 'app/provider',
+      },
+      {
+        contractId: rawConsumerId,
+        type: 'include',
+        role: 'consumer',
+        symbolUid: 'File:src/controller.cpp',
+        symbolRef: { filePath: 'src/controller.cpp', name: 'Map/Base/View.h' },
+        symbolName: 'Map/Base/View.h',
+        confidence: 0.85,
+        meta: { source: 'tree_sitter', includePath: 'Map\\Base\\View.h' },
+        repo: 'app/consumer',
+      },
+    ];
+
+    const result = await syncGroup(config, {
+      extractorOverride: async () => mockContracts,
+      skipWrite: true,
+    });
+
+    const includeLinks = result.crossLinks.filter((l) => l.type === 'include');
+    expect(includeLinks.length).toBeGreaterThanOrEqual(1);
+  });
+
+  it('round-trip: extractor output from two real temp repos produces matching contract-ids', async () => {
+    // Drives the extractor directly (no `syncGroup`) against two on-disk
+    // fixture repos, then hands the StoredContract-shaped output to
+    // syncGroup via extractorOverride. This exercises the real extraction
+    // code + the matching pipeline together.
+    const providerDir = fs.mkdtempSync(path.join(os.tmpdir(), 'gitnexus-include-int-provider-'));
+    const consumerDir = fs.mkdtempSync(path.join(os.tmpdir(), 'gitnexus-include-int-consumer-'));
+    try {
+      fs.mkdirSync(path.join(providerDir, 'shared/api'), { recursive: true });
+      fs.writeFileSync(
+        path.join(providerDir, 'shared/api/client.h'),
+        '#pragma once\nstruct Client {};',
+      );
+      fs.mkdirSync(path.join(consumerDir, 'src'), { recursive: true });
+      fs.writeFileSync(
+        path.join(consumerDir, 'src/main.cpp'),
+        '#include "shared/api/client.h"\nint main(){return 0;}',
+      );
+
+      const extractor = new IncludeExtractor();
+      const providerOutput = await extractor.extract(null, providerDir, {
+        id: 'provider',
+        path: 'app/provider',
+        repoPath: providerDir,
+        storagePath: path.join(providerDir, '.gitnexus'),
+      });
+      const consumerOutput = await extractor.extract(null, consumerDir, {
+        id: 'consumer',
+        path: 'app/consumer',
+        repoPath: consumerDir,
+        storagePath: path.join(consumerDir, '.gitnexus'),
+      });
+
+      const stored: StoredContract[] = [
+        ...providerOutput
+          .filter((c) => c.role === 'provider')
+          .map((c) => ({ ...c, repo: 'app/provider' })),
+        ...consumerOutput
+          .filter((c) => c.role === 'consumer')
+          .map((c) => ({ ...c, repo: 'app/consumer' })),
+      ];
+
+      const config = parseGroupConfig(GROUP_YAML);
+      const result = await syncGroup(config, {
+        extractorOverride: async () => stored,
+        skipWrite: true,
+      });
+
+      const includeLinks = result.crossLinks.filter((l) => l.type === 'include');
+      expect(includeLinks.length).toBeGreaterThanOrEqual(1);
+      expect(includeLinks[0].contractId).toBe('include::shared/api/client.h');
+      expect(includeLinks[0].matchType).toBe('exact');
+    } finally {
+      fs.rmSync(providerDir, { recursive: true, force: true });
+      fs.rmSync(consumerDir, { recursive: true, force: true });
+    }
+  });
+});
diff --git a/gitnexus/test/integration/lbug-close-handle-release.test.ts b/gitnexus/test/integration/lbug-close-handle-release.test.ts
new file mode 100644
index 000000000..c0a3b8758
--- /dev/null
+++ b/gitnexus/test/integration/lbug-close-handle-release.test.ts
@@ -0,0 +1,41 @@
+/**
+ * Integration test: safeClose's Windows post-close handle-release wait.
+ *
+ * On Windows, libuv reports `db.close()` resolved before the kernel has
+ * released the file handle. A subsequent open of the same path can then
+ * race the release and surface "Could not set lock on file". `safeClose`
+ * probes the file with `fs.open` to force the residual lock to surface,
+ * absorbed by the open-time retry in `lbug-config.ts`.
+ */
+import path from 'path';
+import { describe, it } from 'vitest';
+import { createTempDir } from '../helpers/test-db.js';
+
+describe('safeClose — close + reopen does not surface lock errors', () => {
+  it('survives 10 sequential open/close/reopen cycles on the same path', async () => {
+    const tmp = await createTempDir('gitnexus-lbug-close-cycle-');
+    const dbPath = path.join(tmp.dbPath, 'lbug');
+    try {
+      const adapter = await import('../../src/core/lbug/lbug-adapter.js');
+      for (let i = 0; i < 10; i++) {
+        await adapter.initLbug(dbPath);
+        await adapter.closeLbug();
+      }
+    } finally {
+      await tmp.cleanup();
+    }
+  });
+
+  it('safeClose is idempotent — calling twice in a row does not throw', async () => {
+    const tmp = await createTempDir('gitnexus-lbug-idempotent-');
+    const dbPath = path.join(tmp.dbPath, 'lbug');
+    try {
+      const adapter = await import('../../src/core/lbug/lbug-adapter.js');
+      await adapter.initLbug(dbPath);
+      await adapter.closeLbug();
+      await adapter.closeLbug();
+    } finally {
+      await tmp.cleanup();
+    }
+  });
+});
diff --git a/gitnexus/test/integration/lbug-lock-retry.test.ts b/gitnexus/test/integration/lbug-lock-retry.test.ts
index 1279c77fb..b95092e2d 100644
--- a/gitnexus/test/integration/lbug-lock-retry.test.ts
+++ b/gitnexus/test/integration/lbug-lock-retry.test.ts
@@ -14,7 +14,7 @@ import { withTestLbugDB } from '../helpers/test-indexed-db.js';
 
 // Pure-function tests — no DB needed, but grouped here for cohesion
 // with the retry logic they guard.
-import { isDbBusyError } from '../../src/core/lbug/lbug-adapter.js';
+import { isDbBusyError } from '../../src/core/lbug/lbug-config.js';
 
 describe('isDbBusyError', () => {
   it('returns true for "busy" errors (case-insensitive)', () => {
@@ -46,6 +46,18 @@ describe('isDbBusyError', () => {
     expect(isDbBusyError(undefined)).toBe(false);
   });
 
+  // Documented behavior for lock-shaped strings: the matcher is intentionally
+  // broad because in graph-DB contexts these are all transient. If LadybugDB
+  // ever surfaces a non-transient lock-shaped error (e.g., a recovery-time
+  // "lock file missing"), tighten the matcher and add a negative test here
+  // rather than raising the retry budget.
+  it('treats other lock-shaped errors as transient (current intentional behavior)', () => {
+    expect(isDbBusyError(new Error('deadlock detected'))).toBe(true);
+    expect(isDbBusyError(new Error('unlock failed'))).toBe(true);
+    expect(isDbBusyError(new Error('lock contention'))).toBe(true);
+    expect(isDbBusyError(new Error('Could not open lock file'))).toBe(true);
+  });
+
   it('handles non-Error values gracefully', () => {
     expect(isDbBusyError('BUSY error')).toBe(true);
     expect(isDbBusyError(42)).toBe(false);
diff --git a/gitnexus/test/integration/lbug-open-retry.test.ts b/gitnexus/test/integration/lbug-open-retry.test.ts
new file mode 100644
index 000000000..80e68617d
--- /dev/null
+++ b/gitnexus/test/integration/lbug-open-retry.test.ts
@@ -0,0 +1,310 @@
+/**
+ * Integration tests: open-time lock-busy retry in `lbug-config.ts`.
+ *
+ * The lock IO exception raised by `local_file_system.cpp` happens
+ * synchronously inside `new lbug.Database(...)`, before any query is
+ * issued — so `withLbugDb`'s query-time retry cannot see it. These tests
+ * exercise the construction-time retry wrapper directly by stubbing the
+ * `Database` constructor.
+ *
+ * See: docs/plans/2026-05-08-002-fix-windows-lbug-lock-ci-flakes-plan.md
+ */
+import fs from 'fs/promises';
+import os from 'os';
+import path from 'path';
+import { describe, it, expect, beforeEach, afterEach } from 'vitest';
+import {
+  _isTestFixturePathForTest as isTestFixturePath,
+  isDbBusyError,
+  isOpenRetryExhausted,
+  openLbugConnection,
+  waitForWindowsHandleRelease,
+} from '../../src/core/lbug/lbug-config.js';
+
+// ─── Minimal stub of the `lbug` module surface used by openLbugConnection ──
+
+interface StubModuleControl {
+  /** Errors thrown by sequential `new Database(...)` calls. `null` = success. */
+  databaseThrows: Array<Error | null>;
+  /** Number of times the `Database` constructor was invoked. */
+  databaseCallCount: number;
+  /** Number of times `db.close()` was called. */
+  closeCallCount: number;
+}
+
+const makeStubLbug = (control: StubModuleControl) => {
+  class FakeDatabase {
+    constructor(_path: string, ..._rest: unknown[]) {
+      control.databaseCallCount++;
+      const next = control.databaseThrows.shift();
+      if (next instanceof Error) throw next;
+    }
+    async close(): Promise<void> {
+      control.closeCallCount++;
+    }
+  }
+  class FakeConnection {
+    constructor(_db: FakeDatabase) {}
+    async close(): Promise<void> {}
+  }
+  return { Database: FakeDatabase, Connection: FakeConnection } as any;
+};
+
+describe('isDbBusyError', () => {
+  it('matches the documented Windows lock-error wording', () => {
+    expect(isDbBusyError(new Error('Could not set lock on file foo.lbug'))).toBe(true);
+    expect(isDbBusyError(new Error('database is locked'))).toBe(true);
+  });
+  it('does not match unrelated errors', () => {
+    expect(isDbBusyError(new Error('Cypher syntax error'))).toBe(false);
+    expect(isDbBusyError(null)).toBe(false);
+  });
+});
+
+describe('openLbugConnection — open-time lock-busy retry', () => {
+  it('returns a handle when the constructor succeeds on the first try', async () => {
+    const control: StubModuleControl = {
+      databaseThrows: [null],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+    const handle = await openLbugConnection(stub, '/some/path/lbug');
+    expect(handle.db).toBeDefined();
+    expect(handle.conn).toBeDefined();
+    expect(control.databaseCallCount).toBe(1);
+  });
+
+  it('retries on busy/lock errors and succeeds on a later attempt', async () => {
+    const control: StubModuleControl = {
+      databaseThrows: [new Error('Could not set lock on file'), null],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+    const handle = await openLbugConnection(stub, '/some/path/lbug');
+    expect(handle.db).toBeDefined();
+    expect(control.databaseCallCount).toBe(2);
+  });
+
+  it('exhausts the retry budget and rethrows the last error preserving its message', async () => {
+    const lockErr = new Error('Could not set lock on file foo.lbug');
+    const control: StubModuleControl = {
+      // 5 attempts + production paths get no sweep retry, so 5 throws total.
+      databaseThrows: [lockErr, lockErr, lockErr, lockErr, lockErr],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+    await expect(openLbugConnection(stub, '/var/data/non-test/lbug')).rejects.toThrow(
+      'Could not set lock on file foo.lbug',
+    );
+    expect(control.databaseCallCount).toBe(5);
+  });
+
+  it('tags the exhausted error so withLbugDb skips its outer retry', async () => {
+    const lockErr = new Error('Could not set lock on file');
+    const control: StubModuleControl = {
+      databaseThrows: [lockErr, lockErr, lockErr, lockErr, lockErr],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+    let caught: unknown;
+    try {
+      await openLbugConnection(stub, '/var/data/non-test/lbug');
+    } catch (err) {
+      caught = err;
+    }
+    expect(caught).toBeDefined();
+    expect(isOpenRetryExhausted(caught)).toBe(true);
+    expect(isOpenRetryExhausted(new Error('plain error'))).toBe(false);
+    expect(isOpenRetryExhausted(null)).toBe(false);
+    expect(isOpenRetryExhausted(undefined)).toBe(false);
+  });
+
+  it('does not retry non-busy errors', async () => {
+    const syntaxErr = new Error('Cypher syntax error');
+    const control: StubModuleControl = {
+      databaseThrows: [syntaxErr],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+    await expect(openLbugConnection(stub, '/some/path/lbug')).rejects.toThrow(
+      'Cypher syntax error',
+    );
+    expect(control.databaseCallCount).toBe(1);
+  });
+});
+
+describe('openLbugConnection — stale-sidecar sweep (test fixtures only)', () => {
+  let fixtureDir: string;
+  let dbPath: string;
+
+  beforeEach(async () => {
+    fixtureDir = await fs.mkdtemp(path.join(os.tmpdir(), 'gitnexus-lbug-sweep-'));
+    dbPath = path.join(fixtureDir, 'lbug');
+  });
+
+  afterEach(async () => {
+    await fs.rm(fixtureDir, { recursive: true, force: true }).catch(() => {});
+  });
+
+  it('sweeps stale .wal/.lock for a recognized test fixture path and retries once', async () => {
+    await fs.writeFile(dbPath + '.wal', 'stale');
+    await fs.writeFile(dbPath + '.lock', 'stale');
+
+    const lockErr = new Error('Could not set lock on file');
+    const control: StubModuleControl = {
+      // 5 retries throw, then sweep + 1 final attempt succeeds (6 total).
+      databaseThrows: [lockErr, lockErr, lockErr, lockErr, lockErr, null],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+
+    const handle = await openLbugConnection(stub, dbPath);
+    expect(handle.db).toBeDefined();
+    expect(control.databaseCallCount).toBe(6);
+
+    // Sidecars removed by the sweep
+    await expect(fs.access(dbPath + '.wal')).rejects.toThrow();
+    await expect(fs.access(dbPath + '.lock')).rejects.toThrow();
+  });
+
+  it('does not sweep production paths even if they share the prefix', async () => {
+    // A non-tmp dir that *starts* with the prefix must still be rejected.
+    const lockErr = new Error('Could not set lock on file');
+    const control: StubModuleControl = {
+      databaseThrows: [lockErr, lockErr, lockErr, lockErr, lockErr],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+
+    // Path is outside os.tmpdir() so the predicate must reject it.
+    await expect(openLbugConnection(stub, '/var/data/gitnexus-lbug-fake/lbug')).rejects.toThrow(
+      'Could not set lock on file',
+    );
+    expect(control.databaseCallCount).toBe(5); // no sweep retry
+  });
+
+  it('handles missing sidecars gracefully (ENOENT swallowed, retry runs)', async () => {
+    // No .wal or .lock pre-created — sweep ENOENTs both, then succeeds.
+    const lockErr = new Error('Could not set lock on file');
+    const control: StubModuleControl = {
+      databaseThrows: [lockErr, lockErr, lockErr, lockErr, lockErr, null],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+
+    const handle = await openLbugConnection(stub, dbPath);
+    expect(handle.db).toBeDefined();
+    expect(control.databaseCallCount).toBe(6);
+  });
+
+  it('sweep retry that throws a different error preserves the original lock error', async () => {
+    // 5 lock errors, then sweep fires, then post-sweep throws an unrelated
+    // error. The user-actionable signal is "lock retries exhausted" — the
+    // post-sweep error must NOT shadow the original lock message.
+    const lockErr = new Error('Could not set lock on file foo.lbug');
+    const unrelatedErr = new Error('Schema validation error during open');
+    const control: StubModuleControl = {
+      databaseThrows: [lockErr, lockErr, lockErr, lockErr, lockErr, unrelatedErr],
+      databaseCallCount: 0,
+      closeCallCount: 0,
+    };
+    const stub = makeStubLbug(control);
+
+    let caught: Error | undefined;
+    try {
+      await openLbugConnection(stub, dbPath);
+    } catch (err) {
+      caught = err as Error;
+    }
+    expect(caught?.message).toBe('Could not set lock on file foo.lbug');
+    expect(control.databaseCallCount).toBe(6); // sweep retry did fire
+  });
+});
+
+describe('isTestFixturePath — production-safety guard', () => {
+  it('accepts a fixture under os.tmpdir with a recognized prefix on the immediate parent', () => {
+    const tmp = os.tmpdir();
+    expect(isTestFixturePath(path.join(tmp, 'gitnexus-lbug-XXX', 'lbug'))).toBe(true);
+    expect(isTestFixturePath(path.join(tmp, 'gitnexus-test-YYY', 'lbug'))).toBe(true);
+  });
+
+  it('rejects production paths even with a matching prefix', () => {
+    expect(isTestFixturePath('/var/data/gitnexus-lbug-fake/lbug')).toBe(false);
+    expect(isTestFixturePath('/home/user/gitnexus-test-foo/lbug')).toBe(false);
+  });
+
+  it('rejects path traversal attempts that resolve outside tmpdir', () => {
+    const tmp = os.tmpdir();
+    const traversal = path.join(tmp, 'gitnexus-lbug-x', '..', '..', 'etc', 'passwd');
+    expect(isTestFixturePath(traversal)).toBe(false);
+  });
+
+  it('rejects when the immediate parent does not match even if a deeper ancestor does', () => {
+    // Tightening: ancestor walk would have allowed nested paths under
+    // `<tmp>/gitnexus-lbug-x/inner/lbug` to satisfy the predicate. We
+    // require the immediate parent to match.
+    const tmp = os.tmpdir();
+    expect(isTestFixturePath(path.join(tmp, 'gitnexus-lbug-x', 'inner', 'lbug'))).toBe(false);
+  });
+
+  it('handles tmpdir trailing-separator gracefully', () => {
+    // Some Windows TMP configs return a trailing separator; the predicate
+    // strips it before the prefix check so fixtures still match.
+    const tmp = os.tmpdir();
+    const fixture = path.join(tmp, 'gitnexus-lbug-trailing', 'lbug');
+    // Whether or not os.tmpdir() itself has a trailing separator,
+    // the predicate must accept legit fixtures.
+    expect(isTestFixturePath(fixture)).toBe(true);
+  });
+
+  it('rejects unrelated prefixes in tmpdir', () => {
+    const tmp = os.tmpdir();
+    expect(isTestFixturePath(path.join(tmp, 'random-dir', 'lbug'))).toBe(false);
+    expect(isTestFixturePath(path.join(tmp, 'malicious', 'lbug'))).toBe(false);
+  });
+});
+
+describe('waitForWindowsHandleRelease', () => {
+  let fixtureDir: string;
+  let dbPath: string;
+
+  beforeEach(async () => {
+    fixtureDir = await fs.mkdtemp(path.join(os.tmpdir(), 'gitnexus-lbug-probe-'));
+    dbPath = path.join(fixtureDir, 'lbug');
+  });
+
+  afterEach(async () => {
+    await fs.rm(fixtureDir, { recursive: true, force: true }).catch(() => {});
+  });
+
+  it('returns true when the file exists and is openable', async () => {
+    await fs.writeFile(dbPath, 'fake-db-content');
+    const released = await waitForWindowsHandleRelease(dbPath);
+    expect(released).toBe(true);
+  });
+
+  it('returns true when the file does not exist (ENOENT is non-lock)', async () => {
+    // No fs.writeFile — path does not exist. Probe should bail to true,
+    // not retry, since ENOENT is not a lock code.
+    const released = await waitForWindowsHandleRelease(dbPath);
+    expect(released).toBe(true);
+  });
+
+  it('does not leak the file handle when close succeeds', async () => {
+    // Smoke test: 50 sequential probes with a real file. If close were
+    // skipped, fd usage would climb. We rely on test process not OOMing
+    // as the simplest indicator; fd table caps catch egregious leaks.
+    await fs.writeFile(dbPath, 'fake-db-content');
+    for (let i = 0; i < 50; i++) {
+      await waitForWindowsHandleRelease(dbPath);
+    }
+  });
+});
diff --git a/gitnexus/test/integration/resolvers/csharp.test.ts b/gitnexus/test/integration/resolvers/csharp.test.ts
index 613ff7ec2..073b2da16 100644
--- a/gitnexus/test/integration/resolvers/csharp.test.ts
+++ b/gitnexus/test/integration/resolvers/csharp.test.ts
@@ -1440,6 +1440,24 @@ describe('Write access tracking (C#)', () => {
   });
 });
 
+// ---------------------------------------------------------------------------
+// Generic type references: IEntityTypeConfiguration<USER_INFO>, List<USER_INFO>
+// ---------------------------------------------------------------------------
+
+describe('C# generic type-reference tracking', () => {
+  let result: PipelineResult;
+
+  beforeAll(async () => {
+    result = await runPipelineFromRepo(path.join(FIXTURES, 'csharp-generic-type-refs'), () => {});
+  }, 60000);
+
+  it('emits USES edges for generic type arguments', () => {
+    const uses = getRelationships(result, 'USES').filter((e) => e.target === 'USER_INFO');
+    expect(edgeSet(uses)).toContain('UserInfoConfiguration → USER_INFO');
+    expect(edgeSet(uses)).toContain('Load → USER_INFO');
+  });
+});
+
 // ---------------------------------------------------------------------------
 // Call-result variable binding (Phase 9): var user = GetUser(); user.Save()
 // ---------------------------------------------------------------------------
diff --git a/gitnexus/test/integration/resolvers/helpers.ts b/gitnexus/test/integration/resolvers/helpers.ts
index 5368c1715..e92526b19 100644
--- a/gitnexus/test/integration/resolvers/helpers.ts
+++ b/gitnexus/test/integration/resolvers/helpers.ts
@@ -11,6 +11,9 @@ import type { GraphRelationship } from 'gitnexus-shared';
 const LEGACY_RESOLVER_PARITY_EXPECTED_FAILURES: Readonly<Record<string, ReadonlySet<string>>> = {
   csharp: new Set([
     'emits the using-import edge App/Program.cs -> Models/User.cs through the scope-resolution path',
+    // Generic type-argument USES edges are emitted by the registry-primary
+    // resolver only; the legacy DAG path does not synthesize these references.
+    'emits USES edges for generic type arguments',
   ]),
   go: new Set([
     // The legacy DAG path does not resolve method calls when the method is
diff --git a/gitnexus/test/integration/search-core.test.ts b/gitnexus/test/integration/search-core.test.ts
index 2ccc14706..e49169b30 100644
--- a/gitnexus/test/integration/search-core.test.ts
+++ b/gitnexus/test/integration/search-core.test.ts
@@ -19,7 +19,7 @@ withTestLbugDB(
   (_handle) => {
     describe('searchFTSFromLbug — core adapter (no repoId)', () => {
       it('returns ranked results for a matching query', async () => {
-        const results = await searchFTSFromLbug('user authentication', 10);
+        const { results } = await searchFTSFromLbug('user authentication', 10);
 
         expect(results.length).toBeGreaterThan(0);
 
@@ -40,7 +40,7 @@ withTestLbugDB(
       });
 
       it('results are ordered by descending score', async () => {
-        const results = await searchFTSFromLbug('user authentication', 10);
+        const { results } = await searchFTSFromLbug('user authentication', 10);
 
         for (let i = 1; i < results.length; i++) {
           expect(results[i - 1].score).toBeGreaterThanOrEqual(results[i].score);
@@ -48,7 +48,7 @@ withTestLbugDB(
       });
 
       it('auth-related files rank higher than unrelated files', async () => {
-        const results = await searchFTSFromLbug('user authentication', 10);
+        const { results } = await searchFTSFromLbug('user authentication', 10);
         const filePaths = results.map((r) => r.filePath);
 
         expect(filePaths).toContain('src/auth.ts');
@@ -61,7 +61,7 @@ withTestLbugDB(
       });
 
       it('merges scores from multiple node types for the same filePath', async () => {
-        const results = await searchFTSFromLbug('user authentication', 20);
+        const { results } = await searchFTSFromLbug('user authentication', 20);
 
         const authResult = results.find((r) => r.filePath === 'src/auth.ts');
         expect(authResult).toBeDefined();
@@ -73,12 +73,12 @@ withTestLbugDB(
       });
 
       it('respects limit parameter', async () => {
-        const results = await searchFTSFromLbug('user authentication', 2);
+        const { results } = await searchFTSFromLbug('user authentication', 2);
         expect(results.length).toBeLessThanOrEqual(2);
       });
 
       it('returns empty array for a non-matching query', async () => {
-        const results = await searchFTSFromLbug('xyzzyplughtwisty', 10);
+        const { results } = await searchFTSFromLbug('xyzzyplughtwisty', 10);
         expect(results).toEqual([]);
       });
     });
@@ -87,32 +87,32 @@ withTestLbugDB(
 
     describe('unhappy paths', () => {
       it('returns empty array for empty query string', async () => {
-        const results = await searchFTSFromLbug('', 10);
+        const { results } = await searchFTSFromLbug('', 10);
         expect(results).toEqual([]);
       });
 
       it('returns empty array for whitespace-only query', async () => {
-        const results = await searchFTSFromLbug('   ', 10);
+        const { results } = await searchFTSFromLbug('   ', 10);
         expect(results).toEqual([]);
       });
 
       it('handles special characters in query gracefully', async () => {
-        const results = await searchFTSFromLbug('user* OR auth+', 10);
+        const { results } = await searchFTSFromLbug('user* OR auth+', 10);
         expect(Array.isArray(results)).toBe(true);
       });
 
       it('handles limit of 0', async () => {
-        const results = await searchFTSFromLbug('user authentication', 0);
+        const { results } = await searchFTSFromLbug('user authentication', 0);
         expect(results).toEqual([]);
       });
 
       it('handles negative limit gracefully', async () => {
-        const results = await searchFTSFromLbug('user authentication', -1);
+        const { results } = await searchFTSFromLbug('user authentication', -1);
         expect(Array.isArray(results)).toBe(true);
       });
 
       it('handles very large limit', async () => {
-        const results = await searchFTSFromLbug('user authentication', 100000);
+        const { results } = await searchFTSFromLbug('user authentication', 100000);
         expect(results.length).toBeLessThanOrEqual(100000);
         expect(results.length).toBeGreaterThan(0);
       });
diff --git a/gitnexus/test/integration/search-pool.test.ts b/gitnexus/test/integration/search-pool.test.ts
index c0943483b..88128fcfa 100644
--- a/gitnexus/test/integration/search-pool.test.ts
+++ b/gitnexus/test/integration/search-pool.test.ts
@@ -19,7 +19,7 @@ withTestLbugDB(
   (handle) => {
     describe('searchFTSFromLbug — MCP pool adapter (with repoId)', () => {
       it('returns ranked results via pool adapter', async () => {
-        const results = await searchFTSFromLbug('user authentication', 10, handle.repoId);
+        const { results } = await searchFTSFromLbug('user authentication', 10, handle.repoId);
 
         expect(results.length).toBeGreaterThan(0);
 
@@ -35,7 +35,7 @@ withTestLbugDB(
       });
 
       it('results are ordered by descending score via pool adapter', async () => {
-        const results = await searchFTSFromLbug('user authentication', 10, handle.repoId);
+        const { results } = await searchFTSFromLbug('user authentication', 10, handle.repoId);
 
         for (let i = 1; i < results.length; i++) {
           expect(results[i - 1].score).toBeGreaterThanOrEqual(results[i].score);
@@ -43,12 +43,12 @@ withTestLbugDB(
       });
 
       it('returns empty array for non-matching query via pool adapter', async () => {
-        const results = await searchFTSFromLbug('xyzzyplughtwisty', 10, handle.repoId);
+        const { results } = await searchFTSFromLbug('xyzzyplughtwisty', 10, handle.repoId);
         expect(results).toEqual([]);
       });
 
       it('respects limit parameter via pool adapter', async () => {
-        const results = await searchFTSFromLbug('user authentication', 1, handle.repoId);
+        const { results } = await searchFTSFromLbug('user authentication', 1, handle.repoId);
         expect(results.length).toBeLessThanOrEqual(1);
       });
     });
@@ -57,22 +57,22 @@ withTestLbugDB(
 
     describe('unhappy paths', () => {
       it('returns empty array for empty query via pool', async () => {
-        const results = await searchFTSFromLbug('', 10, handle.repoId);
+        const { results } = await searchFTSFromLbug('', 10, handle.repoId);
         expect(results).toEqual([]);
       });
 
       it('returns empty array for whitespace-only query via pool', async () => {
-        const results = await searchFTSFromLbug('   ', 10, handle.repoId);
+        const { results } = await searchFTSFromLbug('   ', 10, handle.repoId);
         expect(results).toEqual([]);
       });
 
       it('handles special characters in query via pool', async () => {
-        const results = await searchFTSFromLbug('user* OR auth+', 10, handle.repoId);
+        const { results } = await searchFTSFromLbug('user* OR auth+', 10, handle.repoId);
         expect(Array.isArray(results)).toBe(true);
       });
 
       it('handles limit of 0 via pool', async () => {
-        const results = await searchFTSFromLbug('user authentication', 0, handle.repoId);
+        const { results } = await searchFTSFromLbug('user authentication', 0, handle.repoId);
         expect(results).toEqual([]);
       });
     });
diff --git a/gitnexus/test/unit/bm25-search.test.ts b/gitnexus/test/unit/bm25-search.test.ts
index 9b232688f..03a591599 100644
--- a/gitnexus/test/unit/bm25-search.test.ts
+++ b/gitnexus/test/unit/bm25-search.test.ts
@@ -42,20 +42,24 @@ describe('BM25 search', () => {
   });
 
   describe('searchFTSFromLbug', () => {
-    it('returns empty array when LadybugDB is not initialized', async () => {
-      // Without LadybugDB init, search should return empty (not crash)
-      const results = await searchFTSFromLbug('test query');
+    it('returns empty results when LadybugDB is not initialized', async () => {
+      // Simulate an uninitialized DB: queryFTS throws instead of returning rows
+      const { queryFTS } = await import('../../src/core/lbug/lbug-adapter.js');
+      vi.mocked(queryFTS).mockRejectedValue(new Error('DB not initialized'));
+
+      const { results, ftsAvailable } = await searchFTSFromLbug('test query');
       expect(Array.isArray(results)).toBe(true);
       expect(results).toHaveLength(0);
+      expect(ftsAvailable).toBe(false);
     });
 
     it('handles empty query', async () => {
-      const results = await searchFTSFromLbug('');
+      const { results } = await searchFTSFromLbug('');
       expect(Array.isArray(results)).toBe(true);
     });
 
     it('accepts custom limit parameter', async () => {
-      const results = await searchFTSFromLbug('test', 5);
+      const { results } = await searchFTSFromLbug('test', 5);
       expect(Array.isArray(results)).toBe(true);
     });
   });
@@ -105,7 +109,7 @@ describe('BM25 search', () => {
         .mockResolvedValueOnce([]) // Method
         .mockResolvedValueOnce([]); // Interface
 
-      const results = await searchFTSFromLbug('queryset');
+      const { results } = await searchFTSFromLbug('queryset');
 
       expect(results).toHaveLength(1);
       expect(results[0].filePath).toBe('src/views.py');
@@ -127,7 +131,7 @@ describe('BM25 search', () => {
         .mockResolvedValueOnce([]) // Method
         .mockResolvedValueOnce([]); // Interface
 
-      const results = await searchFTSFromLbug('model');
+      const { results } = await searchFTSFromLbug('model');
 
       expect(results).toHaveLength(1);
       expect(results[0].score).toBe(8); // 5+3
@@ -147,7 +151,7 @@ describe('BM25 search', () => {
         .mockResolvedValueOnce([]) // Method
         .mockResolvedValueOnce([]); // Interface
 
-      const results = await searchFTSFromLbug('util');
+      const { results } = await searchFTSFromLbug('util');
 
       expect(results).toHaveLength(1);
       expect(results[0].nodeIds).toEqual([]);
@@ -171,7 +175,7 @@ describe('BM25 search', () => {
         .mockResolvedValueOnce([]) // Method
         .mockResolvedValueOnce([]); // Interface
 
-      const results = await searchFTSFromLbug('auth');
+      const { results } = await searchFTSFromLbug('auth');
 
       expect(results).toHaveLength(1);
       // All 3 hits (scores 9+7+4=20) — each from a different table, all top-3
@@ -192,7 +196,7 @@ describe('BM25 search', () => {
         .mockResolvedValueOnce([]) // Method
         .mockResolvedValueOnce([]); // Interface
 
-      const results = await searchFTSFromLbug('fn');
+      const { results } = await searchFTSFromLbug('fn');
 
       expect(results[0].filePath).toBe('src/high.py');
       expect(results[1].filePath).toBe('src/low.py');
@@ -220,7 +224,7 @@ describe('BM25 search', () => {
         return [];
       });
 
-      const results = await searchFTSFromLbug('login', 5, REPO);
+      const { results } = await searchFTSFromLbug('login', 5, REPO);
 
       expect(results).toEqual([
         { filePath: 'src/auth.ts', score: 8, rank: 1, nodeIds: ['func:login'] },
diff --git a/gitnexus/test/unit/calltool-dispatch.test.ts b/gitnexus/test/unit/calltool-dispatch.test.ts
index f8d94d890..45e1b71d2 100644
--- a/gitnexus/test/unit/calltool-dispatch.test.ts
+++ b/gitnexus/test/unit/calltool-dispatch.test.ts
@@ -48,6 +48,7 @@ vi.mock('../../src/storage/repo-manager.js', () => ({
 // tests don't shell out to git.
 vi.mock('../../src/core/git-staleness.js', () => ({
   checkStaleness: vi.fn().mockReturnValue({ isStale: false, commitsBehind: 0 }),
+  checkStalenessAsync: vi.fn().mockResolvedValue({ isStale: false, commitsBehind: 0 }),
   checkCwdMatch: vi.fn().mockResolvedValue({ match: 'none' }),
 }));
 
@@ -61,7 +62,7 @@ vi.mock('../../src/core/platform/capabilities.js', async (importOriginal) => {
 
 // Also mock the search modules to avoid loading onnxruntime
 vi.mock('../../src/core/search/bm25-index.js', () => ({
-  searchFTSFromLbug: vi.fn().mockResolvedValue([]),
+  searchFTSFromLbug: vi.fn().mockResolvedValue({ results: [], ftsAvailable: true }),
 }));
 
 vi.mock('../../src/mcp/core/embedder.js', () => ({
@@ -194,6 +195,27 @@ describe('LocalBackend.callTool', () => {
     expect(result).toHaveProperty('definitions');
   });
 
+  it('includes FTS-unavailable warning when ftsAvailable is false (#1403)', async () => {
+    const { searchFTSFromLbug } = await import('../../src/core/search/bm25-index.js');
+    vi.mocked(searchFTSFromLbug).mockResolvedValueOnce({ results: [], ftsAvailable: false });
+    (executeParameterized as any).mockResolvedValue([]);
+
+    const result = await backend.callTool('query', { query: 'ProcessActivity' });
+
+    expect(result).toHaveProperty('warning');
+    expect((result as any).warning).toMatch(/gitnexus analyze --force/);
+  });
+
+  it('does not include warning when ftsAvailable is true with zero results', async () => {
+    const { searchFTSFromLbug } = await import('../../src/core/search/bm25-index.js');
+    vi.mocked(searchFTSFromLbug).mockResolvedValueOnce({ results: [], ftsAvailable: true });
+    (executeParameterized as any).mockResolvedValue([]);
+
+    const result = await backend.callTool('query', { query: 'nonexistent' });
+
+    expect(result).not.toHaveProperty('warning');
+  });
+
   it('skips vector index query when VECTOR is unsupported by the platform', async () => {
     const cap = _captureLogger();
     platformMocks.isVectorExtensionSupportedByPlatform.mockReturnValue(false);
diff --git a/gitnexus/test/unit/cli-commands.test.ts b/gitnexus/test/unit/cli-commands.test.ts
index a059a37bc..a42afb25c 100644
--- a/gitnexus/test/unit/cli-commands.test.ts
+++ b/gitnexus/test/unit/cli-commands.test.ts
@@ -10,6 +10,9 @@ vi.mock('../../src/cli/mcp.js', () => ({
 vi.mock('../../src/cli/setup.js', () => ({
   setupCommand: vi.fn(),
 }));
+vi.mock('../../src/cli/publish.js', () => ({
+  publishCommand: vi.fn(),
+}));
 
 describe('CLI commands', () => {
   describe('version', () => {
@@ -84,4 +87,11 @@ describe('CLI commands', () => {
       expect(typeof setupCommand).toBe('function');
     });
   });
+
+  describe('publishCommand', () => {
+    it('is a function', async () => {
+      const { publishCommand } = await import('../../src/cli/publish.js');
+      expect(typeof publishCommand).toBe('function');
+    });
+  });
 });
diff --git a/gitnexus/test/unit/cli-index-help.test.ts b/gitnexus/test/unit/cli-index-help.test.ts
index 59109c8d9..889a7473f 100644
--- a/gitnexus/test/unit/cli-index-help.test.ts
+++ b/gitnexus/test/unit/cli-index-help.test.ts
@@ -63,4 +63,17 @@ describe('CLI help surface', () => {
     expect(result.stdout).toContain('--model <model>');
     expect(result.stdout).toContain('--gist');
   });
+
+  it('publish help names the registry, the token env var, and the opt-out behaviour', () => {
+    const result = runHelp('publish');
+
+    expect(result.status).toBe(0);
+    expect(result.stdout).toContain('--id <owner/repo>');
+    expect(result.stdout).toContain('--skip-git');
+    // Discoverability contract: a contributor scanning `--help` must see
+    // (a) which registry this dispatches to, and (b) the env var that
+    // gates the opt-in. Both are part of the no-token contract.
+    expect(result.stdout).toContain('understand-quickly');
+    expect(result.stdout).toContain('UNDERSTAND_QUICKLY_TOKEN');
+  });
 });
diff --git a/gitnexus/test/unit/cpp-ue-preprocessor.test.ts b/gitnexus/test/unit/cpp-ue-preprocessor.test.ts
new file mode 100644
index 000000000..087c5ba2a
--- /dev/null
+++ b/gitnexus/test/unit/cpp-ue-preprocessor.test.ts
@@ -0,0 +1,272 @@
+import { describe, it, expect } from 'vitest';
+import Parser from 'tree-sitter';
+import CPP from 'tree-sitter-cpp';
+import { stripUeMacros } from '../../src/core/ingestion/cpp-ue-preprocessor.js';
+
+describe('stripUeMacros — detection guard', () => {
+  it('returns input unchanged when no UE markers are present', () => {
+    const src = `class Plain {\npublic:\n  int Get() const;\n};`;
+    expect(stripUeMacros(src)).toBe(src);
+  });
+
+  it('returns input unchanged for STL-style code', () => {
+    const src = `#include <vector>\nstd::vector<int> v;`;
+    expect(stripUeMacros(src)).toBe(src);
+  });
+});
+
+describe('stripUeMacros — length preservation', () => {
+  const ueSamples: string[] = [
+    `UCLASS()\nclass BRAWLUI_API UMyClass : public UObject { GENERATED_BODY() public: UFUNCTION() void Run(); };`,
+    `UPROPERTY(EditAnywhere, BlueprintReadOnly, Category = "Combat") int32 Health;`,
+    `USTRUCT(BlueprintType)\nstruct ENGINE_API FMyData { GENERATED_BODY() float Value; };`,
+    `DECLARE_DYNAMIC_MULTICAST_DELEGATE_TwoParams(FMyDelegate, int32, A, FString, B);`,
+    `UE_DEPRECATED(5.0, "Use NewThing instead") void OldThing();`,
+  ];
+
+  for (const src of ueSamples) {
+    it(`preserves byte length: ${src.slice(0, 40).replace(/\n/g, '\\n')}…`, () => {
+      const out = stripUeMacros(src);
+      expect(out.length).toBe(src.length);
+    });
+
+    it(`preserves newline positions: ${src.slice(0, 40).replace(/\n/g, '\\n')}…`, () => {
+      const out = stripUeMacros(src);
+      const inputNewlines: number[] = [];
+      const outputNewlines: number[] = [];
+      for (let i = 0; i < src.length; i++) {
+        if (src.charCodeAt(i) === 0x0a) inputNewlines.push(i);
+        if (out.charCodeAt(i) === 0x0a) outputNewlines.push(i);
+      }
+      expect(outputNewlines).toEqual(inputNewlines);
+    });
+  }
+});
+
+describe('stripUeMacros — macro removal', () => {
+  it('elides UCLASS(...) with arguments', () => {
+    const src = `UCLASS(BlueprintType, Category="Foo")\nclass UFoo {};`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('UCLASS');
+    expect(out).not.toContain('BlueprintType');
+    expect(out).toContain('class UFoo {};');
+  });
+
+  it('elides UCLASS() with empty parens', () => {
+    const src = `UCLASS()\nclass UBar {};`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('UCLASS');
+    expect(out).toContain('class UBar {};');
+  });
+
+  it('elides MODULE_API export macros (BRAWLUI_API style) when paired with a UE marker', () => {
+    const src = `UCLASS()\nclass BRAWLUI_API UMyClass : public UObject {};`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('BRAWLUI_API');
+    expect(out).toContain('class');
+    expect(out).toContain('UMyClass');
+    expect(out).toContain('public UObject');
+  });
+
+  it('elides multiple distinct *_API tokens in same file when UE marker is present', () => {
+    const src = `UCLASS()\nclass CORE_API A {};\nUCLASS()\nclass UMG_API B : public A {};`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('CORE_API');
+    expect(out).not.toContain('UMG_API');
+    expect(out).toContain('class');
+    expect(out).toContain('A {};');
+  });
+
+  it('elides GENERATED_BODY() inside class body', () => {
+    const src = `class UThing { GENERATED_BODY() public: void Foo(); };`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('GENERATED_BODY');
+    expect(out).toContain('public:');
+    expect(out).toContain('void Foo();');
+  });
+
+  it('elides UFUNCTION(...) before method declarations', () => {
+    const src = `class X { UFUNCTION(BlueprintCallable, Server, Reliable) void DoThing(); };`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('UFUNCTION');
+    expect(out).not.toContain('BlueprintCallable');
+    expect(out).toContain('void DoThing();');
+  });
+
+  it('elides UPROPERTY(...) before field declarations', () => {
+    const src = `class X { UPROPERTY(EditAnywhere) int32 Health; };`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('UPROPERTY');
+    expect(out).not.toContain('EditAnywhere');
+    expect(out).toContain('int32 Health;');
+  });
+
+  it('elides DECLARE_DYNAMIC_MULTICAST_DELEGATE_*Params(...)', () => {
+    const src = `DECLARE_DYNAMIC_MULTICAST_DELEGATE_OneParam(FMyDelegate, int32, Value);\nclass X {};`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('DECLARE_DYNAMIC_MULTICAST_DELEGATE');
+    expect(out).not.toContain('FMyDelegate');
+    expect(out).toContain('class X {};');
+  });
+
+  it('elides UE_DEPRECATED(...) before function declarations', () => {
+    const src = `UE_DEPRECATED(5.1, "Reason") void Old();`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('UE_DEPRECATED');
+    expect(out).not.toContain('5.1');
+    expect(out).toContain('void Old();');
+  });
+});
+
+describe('stripUeMacros — non-UE files left alone', () => {
+  it('does NOT strip standalone *_API identifiers when no UE marker is present', () => {
+    const src = `enum class Status { REST_API = 1, HTTP_API = 2, MY_LIB_API = 3 };\nvoid handle(REST_API status);`;
+    expect(stripUeMacros(src)).toBe(src);
+  });
+
+  it('does NOT strip _API tokens in a file that only mentions DECLARE_DELEGATE-like macros from non-UE codebases', () => {
+    const src = `// Custom delegate framework, not UE\n#define DECLARE_HANDLER(x) void x()\nDECLARE_HANDLER(MyHandler);\nint REST_API = 0;`;
+    expect(stripUeMacros(src)).toBe(src);
+  });
+});
+
+describe('stripUeMacros — non-ASCII content preservation', () => {
+  it('leaves non-ASCII content outside elided ranges intact and at the same .length offset', () => {
+    const src = `// Comment with non-ASCII: café résumé naïve\nUCLASS()\nclass UMyClass : public UObject\n{\n  GENERATED_BODY()\n  // Trailing: 日本語 αβγ\n};`;
+    const out = stripUeMacros(src);
+    expect(out.length).toBe(src.length);
+    expect(out).toContain('café résumé naïve');
+    expect(out).toContain('日本語 αβγ');
+    expect(out).toContain('class UMyClass : public UObject');
+    expect(out).not.toContain('UCLASS');
+    expect(out).not.toContain('GENERATED_BODY');
+  });
+
+  it('preserves newline positions when the file contains non-ASCII characters', () => {
+    const src = `// café\nUPROPERTY()\nint32 Health;\n// résumé\nUFUNCTION()\nvoid Run();`;
+    const out = stripUeMacros(src);
+    const inputNewlines: number[] = [];
+    const outputNewlines: number[] = [];
+    for (let i = 0; i < src.length; i++) {
+      if (src.charCodeAt(i) === 0x0a) inputNewlines.push(i);
+      if (out.charCodeAt(i) === 0x0a) outputNewlines.push(i);
+    }
+    expect(outputNewlines).toEqual(inputNewlines);
+  });
+});
+
+describe('stripUeMacros — false-positive guards', () => {
+  it('does NOT strip identifiers that merely contain UCLASS as a substring', () => {
+    const src = `void NotUCLASSAtAll(); int MyUCLASS = 0;`;
+    const out = stripUeMacros(src);
+    expect(out).toBe(src);
+  });
+
+  it('does NOT strip _API substrings inside larger identifiers', () => {
+    const src = `class MY_APIName {};\nint not_my_API_thing = 0;`;
+    const out = stripUeMacros(src);
+    expect(out).toContain('MY_APIName');
+    expect(out).toContain('not_my_API_thing');
+  });
+
+  it('does not eat parens balanced inside string literals', () => {
+    const src = `UFUNCTION(meta=(DisplayName="Foo (Bar)")) void Z();`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('UFUNCTION');
+    expect(out).not.toContain('DisplayName');
+    expect(out).toContain('void Z();');
+  });
+
+  it('handles UCLASS with deeply nested parens in arguments', () => {
+    const src = `UCLASS(meta=(Categories=("A.B", "C.D")), Within=Foo) class UDeep {};`;
+    const out = stripUeMacros(src);
+    expect(out).not.toContain('UCLASS');
+    expect(out).not.toContain('Categories');
+    expect(out).toContain('class UDeep {};');
+  });
+
+  it('leaves Qt macros alone (only UE markers stripped)', () => {
+    const src = `class QFoo { Q_OBJECT public: void Bar(); };`;
+    const out = stripUeMacros(src);
+    expect(out).toContain('Q_OBJECT');
+  });
+});
+
+describe('stripUeMacros — class-name extraction sanity', () => {
+  it('after stripping, "class UMyClass" appears immediately after "class "', () => {
+    const src = `UCLASS(BlueprintType)\nclass BRAWLUI_API UMyClass : public UObject\n{\n  GENERATED_BODY()\n};`;
+    const out = stripUeMacros(src);
+    const classIdx = out.indexOf('class ');
+    expect(classIdx).toBeGreaterThanOrEqual(0);
+    const tail = out.slice(classIdx + 'class '.length).trimStart();
+    expect(tail.startsWith('UMyClass')).toBe(true);
+  });
+});
+
+describe('stripUeMacros — tree-sitter extraction (end-to-end)', () => {
+  /**
+   * Walk the parse tree and return the captured class name(s). Works against
+   * the actual tree-sitter-cpp grammar so this is a true integration check
+   * for the core PR claim: the indexer now sees `UMyClass`, not `BRAWLUI_API`.
+   */
+  function extractClassNames(source: string): string[] {
+    const parser = new Parser();
+    parser.setLanguage(CPP as unknown as Parser.Language);
+    const tree = parser.parse(source);
+    const names: string[] = [];
+    const stack: Parser.SyntaxNode[] = [tree.rootNode];
+    while (stack.length > 0) {
+      const node = stack.pop()!;
+      if (node.type === 'class_specifier' || node.type === 'struct_specifier') {
+        const nameNode = node.childForFieldName('name');
+        if (nameNode) names.push(nameNode.text);
+      }
+      for (let i = node.namedChildCount - 1; i >= 0; i--) {
+        const child = node.namedChild(i);
+        if (child) stack.push(child);
+      }
+    }
+    return names;
+  }
+
+  it('tree-sitter-cpp captures UMyClass as the class name (not BRAWLUI_API)', () => {
+    const src = `UCLASS(BlueprintType)\nclass BRAWLUI_API UMyClass : public UObject\n{\n  GENERATED_BODY()\n public:\n  UFUNCTION()\n  void Run();\n};`;
+    const out = stripUeMacros(src);
+    const names = extractClassNames(out);
+    expect(names).toContain('UMyClass');
+    expect(names).not.toContain('BRAWLUI_API');
+  });
+
+  it('tree-sitter-cpp captures struct name correctly through USTRUCT + MODULE_API', () => {
+    const src = `USTRUCT(BlueprintType)\nstruct ENGINE_API FMyData : public FBase\n{\n  GENERATED_BODY()\n  float Value;\n};`;
+    const out = stripUeMacros(src);
+    const names = extractClassNames(out);
+    expect(names).toContain('FMyData');
+    expect(names).not.toContain('ENGINE_API');
+  });
+
+  it('tree-sitter-cpp source positions are preserved across stripping (line numbers match)', () => {
+    const src = `UCLASS()\nclass BRAWLUI_API UMyClass : public UObject\n{\n  GENERATED_BODY()\n public:\n  void Run();\n};`;
+    const out = stripUeMacros(src);
+    const parser = new Parser();
+    parser.setLanguage(CPP as unknown as Parser.Language);
+    const tree = parser.parse(out);
+    const stack: Parser.SyntaxNode[] = [tree.rootNode];
+    let runLine: number | undefined;
+    while (stack.length > 0) {
+      const node = stack.pop()!;
+      if (node.type === 'function_declarator') {
+        const declarator = node.childForFieldName('declarator');
+        if (declarator?.text === 'Run') {
+          runLine = node.startPosition.row;
+          break;
+        }
+      }
+      for (let i = node.namedChildCount - 1; i >= 0; i--) {
+        const child = node.namedChild(i);
+        if (child) stack.push(child);
+      }
+    }
+    expect(runLine).toBe(5); // 0-indexed: "void Run();" is on line 6 (index 5)
+  });
+});
diff --git a/gitnexus/test/unit/group/config-parser.test.ts b/gitnexus/test/unit/group/config-parser.test.ts
index e1d3b540f..22bb2ad26 100644
--- a/gitnexus/test/unit/group/config-parser.test.ts
+++ b/gitnexus/test/unit/group/config-parser.test.ts
@@ -75,6 +75,62 @@ repos:
     expect(config.detect.thrift).toBe(true);
   });
 
+  // PR #1156 Codex follow-up: include extraction is opt-in. Existing
+  // group.yaml files that do not declare `detect.includes` must not gain
+  // a wave of new include::* contracts on the next sync after upgrade.
+  describe('detect.includes opt-in default', () => {
+    it('defaults includes detection to false when detect block omits it', () => {
+      const minimal = `
+version: 1
+name: test
+repos:
+  app: my-app
+`;
+      const config = parseGroupConfig(minimal);
+      expect(config.detect.includes).toBe(false);
+    });
+
+    it('defaults includes detection to false when detect block is present but omits the key', () => {
+      const yaml = `
+version: 1
+name: test
+repos:
+  app: my-app
+detect:
+  http: true
+  grpc: false
+`;
+      const config = parseGroupConfig(yaml);
+      expect(config.detect.includes).toBe(false);
+    });
+
+    it('honors explicit detect.includes: true (opt-in works)', () => {
+      const yaml = `
+version: 1
+name: test
+repos:
+  app: my-app
+detect:
+  includes: true
+`;
+      const config = parseGroupConfig(yaml);
+      expect(config.detect.includes).toBe(true);
+    });
+
+    it('honors explicit detect.includes: false', () => {
+      const yaml = `
+version: 1
+name: test
+repos:
+  app: my-app
+detect:
+  includes: false
+`;
+      const config = parseGroupConfig(yaml);
+      expect(config.detect.includes).toBe(false);
+    });
+  });
+
   it('parses thrift manifest links', () => {
     const yaml = `
 version: 1
diff --git a/gitnexus/test/unit/group/cross-impact-phase2-timeout.test.ts b/gitnexus/test/unit/group/cross-impact-phase2-timeout.test.ts
new file mode 100644
index 000000000..90b8a80a2
--- /dev/null
+++ b/gitnexus/test/unit/group/cross-impact-phase2-timeout.test.ts
@@ -0,0 +1,121 @@
+/**
+ * Phase-2 fanout timeout regression test.
+ *
+ * Codex adversarial review on PR #1331 surfaced that `validateGroupImpactParams`
+ * clamps `timeoutMs` and `safeLocalImpact` enforces it on the local leg, but
+ * the Phase-2 cross-repo fanout (`cross-impact.ts:521-526`) awaits each
+ * `port.impactByUid(...)` call without a per-call timeout. A single hung
+ * neighbor pins the request indefinitely; multiple slow neighbors compound
+ * past the clamped budget because each starts before `Date.now() > deadline`.
+ *
+ * This test pins the contract of the mitigation: a `safeNeighborImpact`
+ * helper that races `port.impactByUid` against a remaining-budget timer
+ * and returns `{ value: null, timedOut: true }` when the call cannot
+ * complete in time.
+ *
+ * Direct import + named symbol so this is a real regression net — no
+ * `??`-fallback or dynamic-import dance (the U8 false-green pattern).
+ */
+import { describe, expect, it } from 'vitest';
+import { safeNeighborImpact } from '../../../src/core/group/cross-impact.js';
+import type { GroupToolPort } from '../../../src/core/group/service.js';
+
+const minimalOpts = {
+  maxDepth: 3,
+  relationTypes: [] as string[],
+  minConfidence: 0,
+  includeTests: false,
+};
+
+function makePort(impactByUid: GroupToolPort['impactByUid']): GroupToolPort {
+  return {
+    resolveRepo: async () => {
+      throw new Error('not used');
+    },
+    impact: async () => {
+      throw new Error('not used');
+    },
+    query: async () => {
+      throw new Error('not used');
+    },
+    context: async () => {
+      throw new Error('not used');
+    },
+    impactByUid,
+  };
+}
+
+describe('safeNeighborImpact — Phase-2 fanout per-call timeout', () => {
+  it('returns timedOut=true when impactByUid never resolves, within ~remainingMs', async () => {
+    // Hung neighbor: the promise never resolves. Without the timeout wrap
+    // this would hang the test runner.
+    const port = makePort(() => new Promise(() => {}));
+    const start = performance.now();
+    const result = await safeNeighborImpact(port, 'repo-id', 'uid:1', 'upstream', minimalOpts, 150);
+    const elapsedMs = performance.now() - start;
+    expect(result.timedOut).toBe(true);
+    expect(result.value).toBeNull();
+    // Allow generous slack for slow CI; the contract is "bounded", not
+    // "exactly remainingMs". A regression that drops the timeout entirely
+    // would hang far past 1500ms; a regression that uses the wrong unit
+    // (seconds vs ms) would fire much faster.
+    expect(elapsedMs).toBeGreaterThanOrEqual(140);
+    expect(elapsedMs).toBeLessThan(1500);
+  });
+
+  it('returns the resolved value and timedOut=false on a fast happy path', async () => {
+    const fakeFan = { byDepth: { 1: [{ id: 'u1' }] } };
+    const port = makePort(async () => fakeFan);
+    const result = await safeNeighborImpact(
+      port,
+      'repo-id',
+      'uid:1',
+      'upstream',
+      minimalOpts,
+      1000,
+    );
+    expect(result.timedOut).toBe(false);
+    expect(result.value).toBe(fakeFan);
+  });
+
+  it('returns timedOut=true immediately when remainingMs is 0 and the call still hangs', async () => {
+    // Defensive: even if the caller passes 0, the helper must not block.
+    const port = makePort(() => new Promise(() => {}));
+    const start = performance.now();
+    const result = await safeNeighborImpact(port, 'repo-id', 'uid:1', 'upstream', minimalOpts, 0);
+    const elapsedMs = performance.now() - start;
+    expect(result.timedOut).toBe(true);
+    expect(result.value).toBeNull();
+    // 0ms timeout fires on the next tick — should be well under 50ms even on slow CI.
+    expect(elapsedMs).toBeLessThan(50);
+  });
+
+  it('does not compound across calls — two hung neighbors complete within ~2× remainingMs total', async () => {
+    // The contract is per-call timeout. Two sequential hung calls should
+    // total ~2× remainingMs, not (numNeighbors × remainingMs² / 2) or
+    // anything compounding. A regression that shares one timer across
+    // calls would pass the first test but fail this one.
+    const port = makePort(() => new Promise(() => {}));
+    const start = performance.now();
+    const r1 = await safeNeighborImpact(port, 'repo', 'u1', 'upstream', minimalOpts, 100);
+    const r2 = await safeNeighborImpact(port, 'repo', 'u2', 'upstream', minimalOpts, 100);
+    const elapsedMs = performance.now() - start;
+    expect(r1.timedOut).toBe(true);
+    expect(r2.timedOut).toBe(true);
+    expect(elapsedMs).toBeGreaterThanOrEqual(180);
+    expect(elapsedMs).toBeLessThan(1000);
+  });
+
+  it('propagates an immediate rejection from impactByUid as timedOut=false with null value', async () => {
+    // If the port itself rejects (rather than hangs), the helper should
+    // surface that as a non-timeout failure — the existing fanout block
+    // already handles `if (fan == null)` truncation, so returning null
+    // here keeps that path intact.
+    const port = makePort(async () => {
+      throw new Error('connection refused');
+    });
+    const result = await safeNeighborImpact(port, 'repo', 'u1', 'upstream', minimalOpts, 1000);
+    expect(result.timedOut).toBe(false);
+    expect(result.value).toBeNull();
+  });
+});
diff --git a/gitnexus/test/unit/group/include-extractor.test.ts b/gitnexus/test/unit/group/include-extractor.test.ts
new file mode 100644
index 000000000..3956cd5f2
--- /dev/null
+++ b/gitnexus/test/unit/group/include-extractor.test.ts
@@ -0,0 +1,563 @@
+import { describe, it, expect, beforeEach, afterEach } from 'vitest';
+import * as fs from 'node:fs';
+import * as path from 'node:path';
+import * as os from 'node:os';
+import { IncludeExtractor } from '../../../src/core/group/extractors/include-extractor.js';
+import type { RepoHandle } from '../../../src/core/group/types.js';
+import { normalizeContractId } from '../../../src/core/group/matching.js';
+
+describe('IncludeExtractor', () => {
+  let tmpDir: string;
+  let extractor: IncludeExtractor;
+
+  beforeEach(() => {
+    tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'gitnexus-include-'));
+    extractor = new IncludeExtractor();
+  });
+
+  afterEach(() => {
+    fs.rmSync(tmpDir, { recursive: true, force: true });
+  });
+
+  function writeFile(relPath: string, content: string): void {
+    const full = path.join(tmpDir, relPath);
+    fs.mkdirSync(path.dirname(full), { recursive: true });
+    fs.writeFileSync(full, content);
+  }
+
+  const makeRepo = (repoPath: string): RepoHandle => ({
+    id: 'test-repo',
+    path: 'test/app',
+    repoPath,
+    storagePath: path.join(repoPath, '.gitnexus'),
+  });
+
+  // ---- Provider detection ----
+
+  describe('provider extraction', () => {
+    it('registers .h files as providers', async () => {
+      writeFile('map/base/view.h', '#pragma once\nclass View {};');
+      writeFile('map/base/types.h', '#pragma once\nstruct Point {};');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const providers = contracts.filter((c) => c.role === 'provider');
+
+      expect(providers).toHaveLength(2);
+      const ids = providers.map((p) => p.contractId).sort();
+      expect(ids).toEqual(['include::map/base/types.h', 'include::map/base/view.h']);
+      expect(providers[0].type).toBe('include');
+      expect(providers[0].confidence).toBeGreaterThanOrEqual(0.95);
+    });
+
+    it('registers .hpp files as providers', async () => {
+      writeFile('utils/helper.hpp', '#pragma once\ntemplate<class T> T id(T x) { return x; }');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const providers = contracts.filter((c) => c.role === 'provider');
+
+      expect(providers).toHaveLength(1);
+      expect(providers[0].contractId).toBe('include::utils/helper.hpp');
+    });
+
+    it('does not register .cpp files as providers', async () => {
+      writeFile('src/main.cpp', 'int main() { return 0; }');
+      writeFile('src/utils.h', '#pragma once');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const providers = contracts.filter((c) => c.role === 'provider');
+
+      expect(providers).toHaveLength(1);
+      expect(providers[0].contractId).toBe('include::src/utils.h');
+    });
+  });
+
+  // ---- Consumer detection ----
+
+  describe('consumer extraction', () => {
+    it('emits unresolved includes as consumers', async () => {
+      writeFile(
+        'src/main.cpp',
+        `#include "map/base/view.h"
+#include "map/base/types.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers).toHaveLength(2);
+      const ids = consumers.map((c) => c.contractId).sort();
+      expect(ids).toEqual(['include::map/base/types.h', 'include::map/base/view.h']);
+      expect(consumers[0].type).toBe('include');
+      expect(consumers[0].confidence).toBe(0.85);
+    });
+
+    it('skips locally resolved includes', async () => {
+      writeFile('map/base/view.h', '#pragma once\nclass View {};');
+      writeFile(
+        'src/main.cpp',
+        `#include "map/base/view.h"
+#include "external/lib.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      // Only external/lib.h should be a consumer — map/base/view.h resolves locally
+      expect(consumers).toHaveLength(1);
+      expect(consumers[0].contractId).toBe('include::external/lib.h');
+    });
+
+    it('skips angle-bracket includes', async () => {
+      writeFile(
+        'src/main.cpp',
+        `#include <stdio.h>
+#include <vector>
+#include "app/interface.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers).toHaveLength(1);
+      expect(consumers[0].contractId).toBe('include::app/interface.h');
+    });
+
+    it('skips well-known system headers in quotes', async () => {
+      writeFile(
+        'src/main.cpp',
+        `#include "stdio.h"
+#include "stdlib.h"
+#include "app/config.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers).toHaveLength(1);
+      expect(consumers[0].contractId).toBe('include::app/config.h');
+    });
+
+    it('skips system path prefixes', async () => {
+      writeFile(
+        'src/main.c',
+        `#include "sys/types.h"
+#include "linux/input.h"
+#include "mylib/types.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers).toHaveLength(1);
+      expect(consumers[0].contractId).toBe('include::mylib/types.h');
+    });
+  });
+
+  // ---- Cross-repo matching scenario ----
+
+  describe('cross-repo matching', () => {
+    it('provider and consumer produce matching contractIds', async () => {
+      // Simulate provider repo (header-only)
+      const providerDir = fs.mkdtempSync(path.join(os.tmpdir(), 'gitnexus-include-provider-'));
+      const providerFile = path.join(providerDir, 'map/base/dice_map_view.h');
+      fs.mkdirSync(path.dirname(providerFile), { recursive: true });
+      fs.writeFileSync(providerFile, '#pragma once\nclass DiceMapView {};');
+
+      // Simulate consumer repo
+      const consumerDir = fs.mkdtempSync(path.join(os.tmpdir(), 'gitnexus-include-consumer-'));
+      const consumerFile = path.join(consumerDir, 'src/controller.cpp');
+      fs.mkdirSync(path.dirname(consumerFile), { recursive: true });
+      fs.writeFileSync(consumerFile, '#include "map/base/dice_map_view.h"\nvoid init() {}');
+
+      try {
+        const providerContracts = await extractor.extract(null, providerDir, makeRepo(providerDir));
+        const consumerContracts = await extractor.extract(null, consumerDir, makeRepo(consumerDir));
+
+        const providers = providerContracts.filter((c) => c.role === 'provider');
+        const consumers = consumerContracts.filter((c) => c.role === 'consumer');
+
+        expect(providers.length).toBeGreaterThanOrEqual(1);
+        expect(consumers.length).toBeGreaterThanOrEqual(1);
+
+        const providerIds = new Set(providers.map((p) => normalizeContractId(p.contractId)));
+        const consumerIds = consumers.map((c) => normalizeContractId(c.contractId));
+
+        // The consumer's include path should match a provider's file path
+        expect(providerIds.has(consumerIds[0])).toBe(true);
+      } finally {
+        fs.rmSync(providerDir, { recursive: true, force: true });
+        fs.rmSync(consumerDir, { recursive: true, force: true });
+      }
+    });
+  });
+
+  // ---- Review finding #4: suffixResolve ambiguity ----
+
+  describe('finding #4: suffix-ambiguity does not silently suppress cross-repo include', () => {
+    it('emits a cross-repo contract when the include path does not match any local file (even if a shorter suffix does)', async () => {
+      // local repo has `internal/api.h` but NOT `ext/api.h`
+      writeFile('internal/api.h', '#pragma once');
+      writeFile(
+        'src/main.cpp',
+        `#include "ext/api.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      // Previously suffixResolve would match `api.h` against `internal/api.h`
+      // and drop the cross-repo contract. After finding #4 fix, we only
+      // accept exact full-path matches — so `ext/api.h` must still be
+      // emitted as a consumer contract.
+      expect(consumers).toHaveLength(1);
+      expect(consumers[0].contractId).toBe('include::ext/api.h');
+    });
+
+    it('still suppresses a local include when the FULL path matches', async () => {
+      writeFile('ext/api.h', '#pragma once');
+      writeFile('src/main.cpp', '#include "ext/api.h"\nint main(){return 0;}');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers).toHaveLength(0);
+    });
+
+    it('resolves locally when include omits extension and a matching .h exists', async () => {
+      writeFile('foo/bar.h', '#pragma once');
+      writeFile('src/main.cpp', '#include "foo/bar"\nint main(){return 0;}');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers).toHaveLength(0);
+    });
+  });
+
+  // ---- Review finding #5: regex fallback must strip block comments ----
+
+  describe('finding #5: regex fallback ignores block-commented includes', () => {
+    it('does not emit a contract for an #include inside /* ... */', async () => {
+      // Force regex fallback by producing a file larger than tree-sitter's
+      // 32 KB hard cap. The include we care about lives inside a block
+      // comment that spans the file.
+      const filler = 'int dummy_' + 'x'.repeat(32) + ' = 0;\n'.repeat(1200);
+      const content = `/*
+ * Historical include, kept for reference only:
+ * #include "legacy/old-api.h"
+ */
+${filler}
+#include "real/api.h"
+int main(){return 0;}`;
+      writeFile('src/huge.cpp', content);
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+      const ids = consumers.map((c) => c.contractId);
+
+      // The live include should appear; the commented-out one must NOT.
+      expect(ids).toContain('include::real/api.h');
+      expect(ids).not.toContain('include::legacy/old-api.h');
+    });
+  });
+
+  // ---- Review finding #6: meta.source must reflect which extraction path ran ----
+
+  describe('finding #6: meta.source reflects extraction path', () => {
+    it('stamps `tree_sitter` on contracts produced via AST walking', async () => {
+      writeFile('src/main.cpp', '#include "app/small.h"\nint main(){return 0;}');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers).toHaveLength(1);
+      expect((consumers[0].meta as { source?: string } | undefined)?.source).toBe('tree_sitter');
+    });
+
+    it('meta.source is one of the two documented values (tree_sitter | regex_fallback)', async () => {
+      // Regex fallback is a defensive branch that only fires if
+      // parser.setLanguage() or parser.parse() throws. In practice
+      // tree-sitter-c/cpp handles realistic inputs, so we only assert
+      // the meta.source contract: it is always present and always one of
+      // the two documented values. This guards against future regressions
+      // that might hard-code the wrong string.
+      writeFile('src/main.cpp', '#include "ext/whatever.h"\nint main(){return 0;}');
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumer = contracts.find((c) => c.role === 'consumer');
+      expect(consumer).toBeDefined();
+      const src = (consumer?.meta as { source?: string } | undefined)?.source;
+      expect(['tree_sitter', 'regex_fallback']).toContain(src);
+    });
+  });
+
+  // ---- Review finding #3: provider id collision on case-sensitive FS ----
+
+  describe('finding #3: case-folding is documented and deterministic', () => {
+    it('collapses `Foo.h` and `foo.h` onto the same provider contract-id (documented trade-off)', async () => {
+      writeFile('Foo.h', '#pragma once\n// Capital Foo');
+      // On case-insensitive filesystems (macOS default) the second writeFile
+      // will overwrite the first, so we only create this when distinct files
+      // can coexist (case-sensitive FS, e.g. Linux CI).
+      try {
+        fs.writeFileSync(path.join(tmpDir, 'foo.h'), '#pragma once\n// lowercase foo');
+      } catch {
+        // Ignore — some FS won't allow both names to coexist.
+      }
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const providers = contracts.filter((c) => c.role === 'provider');
+      const ids = providers.map((p) => p.contractId);
+
+      // Both files (if they coexist) must normalize to the same id.
+      // dedupe() keeps only one; caller code must be aware of this.
+      expect(ids).toContain('include::foo.h');
+      // Never see a mixed-case contract-id leak out.
+      expect(ids.every((id) => id === id.toLowerCase())).toBe(true);
+    });
+  });
+
+  // ---- Deduplication ----
+
+  describe('deduplication', () => {
+    it('deduplicates same include from multiple source files', async () => {
+      writeFile('src/a.cpp', '#include "ext/api.h"\nvoid a() {}');
+      writeFile('src/b.cpp', '#include "ext/api.h"\nvoid b() {}');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      // Both files include "ext/api.h" — each should produce a separate
+      // consumer contract (different symbolRef.filePath)
+      expect(consumers).toHaveLength(2);
+      const files = consumers.map((c) => c.symbolRef.filePath).sort();
+      expect(files).toEqual(['src/a.cpp', 'src/b.cpp']);
+    });
+  });
+
+  // ---- normalizeContractId ----
+
+  describe('normalizeContractId for include', () => {
+    it('lowercases the path', () => {
+      expect(normalizeContractId('include::Map/Base/Foo.h')).toBe('include::map/base/foo.h');
+    });
+
+    it('normalizes backslashes', () => {
+      expect(normalizeContractId('include::map\\base\\foo.h')).toBe('include::map/base/foo.h');
+    });
+
+    it('strips leading ./', () => {
+      expect(normalizeContractId('include::./foo.h')).toBe('include::foo.h');
+    });
+
+    it('collapses consecutive slashes', () => {
+      expect(normalizeContractId('include::map//base///foo.h')).toBe('include::map/base/foo.h');
+    });
+  });
+
+  // ---- PR #1156 follow-up: `../` relative includes ----
+
+  describe('follow-up: `../` relative includes are skipped', () => {
+    it('does not emit a consumer contract for `#include "../foo.h"`', async () => {
+      // Producer: a header that exists locally but only via parent reference
+      writeFile('include/foo.h', '#pragma once');
+      writeFile(
+        'src/sub/main.cpp',
+        `#include "../../include/foo.h"
+#include "real/cross_repo.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      // Only `real/cross_repo.h` should remain — the `..`-prefixed include
+      // is intra-repo noise that no provider can ever satisfy.
+      expect(consumers.map((c) => c.contractId)).toEqual(['include::real/cross_repo.h']);
+    });
+
+    it('skips backslash-form `..\\` for completeness', async () => {
+      writeFile(
+        'src/main.cpp',
+        `#include "..\\\\sibling\\\\foo.h"
+#include "remote/header.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      const ids = consumers.map((c) => c.contractId);
+      expect(ids).toContain('include::remote/header.h');
+      expect(ids.some((id) => id.includes('..'))).toBe(false);
+    });
+  });
+
+  // ---- PR #1156 follow-up: macro-style includes ----
+
+  describe('follow-up: macro-style #include emits no consumer contract', () => {
+    it('does not emit a consumer contract for `#include PLATFORM_HEADER` (no separator, no dot)', async () => {
+      // `#include PLATFORM_HEADER` parses under tree-sitter as an identifier
+      // node, slips past the existing system-header / `..` filters, and used
+      // to leak through as a permanently orphaned consumer contract because
+      // no file is ever named `PLATFORM_HEADER`. Verify the macro guard
+      // suppresses it while preserving the real cross-repo include.
+      writeFile(
+        'src/main.cpp',
+        `#include PLATFORM_HEADER
+#include "real/api.h"
+int main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers.map((c) => c.contractId)).toEqual(['include::real/api.h']);
+    });
+
+    it('skips multiple macro identifiers in the same translation unit', async () => {
+      writeFile(
+        'src/cfg.cpp',
+        `#include CONFIG_HEADER
+#include PLATFORM_HEADER
+#include ASSERT_H_
+int main(){return 0;}`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumers = contracts.filter((c) => c.role === 'consumer');
+
+      expect(consumers).toHaveLength(0);
+    });
+  });
+
+  // ---- PR #1156 follow-up: graph provider absolute paths ----
+
+  describe('follow-up: extractProvidersGraph strips repo root from absolute paths', () => {
+    it('produces repo-relative contract IDs when the graph returns absolute paths', async () => {
+      writeFile('map/base/view.h', '#pragma once\nclass View {};');
+      writeFile('utils/types.hpp', '#pragma once');
+
+      // Stub the Cypher executor to return absolute paths the way
+      // gitnexus analyze actually persists them.
+      const absolute1 = path.join(tmpDir, 'map/base/view.h');
+      const absolute2 = path.join(tmpDir, 'utils/types.hpp');
+      const stubDb = async () => [
+        { filePath: absolute1, fileId: 'File:abs:1' },
+        { filePath: absolute2, fileId: 'File:abs:2' },
+      ];
+
+      const contracts = await extractor.extract(stubDb, tmpDir, makeRepo(tmpDir));
+      const providers = contracts.filter((c) => c.role === 'provider');
+
+      const ids = providers.map((p) => p.contractId).sort();
+      expect(ids).toEqual(['include::map/base/view.h', 'include::utils/types.hpp']);
+      expect(providers.every((p) => p.meta?.source === 'graph')).toBe(true);
+    });
+
+    it('drops graph rows whose path resolves outside the repo root', async () => {
+      writeFile('local/header.h', '#pragma once');
+      const absoluteLocal = path.join(tmpDir, 'local/header.h');
+      const stubDb = async () => [
+        { filePath: absoluteLocal, fileId: 'File:1' },
+        // Stale absolute path from a different machine — must be skipped.
+        { filePath: '/some/other/repo/foreign.h', fileId: 'File:2' },
+      ];
+
+      const contracts = await extractor.extract(stubDb, tmpDir, makeRepo(tmpDir));
+      const providers = contracts.filter((c) => c.role === 'provider');
+
+      expect(providers.map((p) => p.contractId)).toEqual(['include::local/header.h']);
+    });
+  });
+
+  // ---- PR #1156 Codex follow-up: discovery aligned with ingestion ----
+
+  describe('follow-up: file discovery honors createIgnoreFilter and getMaxFileSizeBytes', () => {
+    it('does not emit a provider contract for a header excluded by .gitignore', async () => {
+      writeFile('.gitignore', 'vendor-headers/\n');
+      writeFile('vendor-headers/blocked.h', '#pragma once');
+      writeFile('src/wanted.h', '#pragma once');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const providerIds = contracts.filter((c) => c.role === 'provider').map((p) => p.contractId);
+
+      expect(providerIds).toContain('include::src/wanted.h');
+      expect(providerIds).not.toContain('include::vendor-headers/blocked.h');
+    });
+
+    it('does not emit a provider contract for a header excluded by .gitnexusignore', async () => {
+      writeFile('.gitnexusignore', 'legacy/\n');
+      writeFile('legacy/old.h', '#pragma once');
+      writeFile('src/current.h', '#pragma once');
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const providerIds = contracts.filter((c) => c.role === 'provider').map((p) => p.contractId);
+
+      expect(providerIds).toContain('include::src/current.h');
+      expect(providerIds).not.toContain('include::legacy/old.h');
+    });
+
+    it('does not parse #include directives in a source file excluded by .gitignore', async () => {
+      // The ignored source file references a header that would otherwise be
+      // a cross-repo consumer. After alignment, the ignored file is invisible
+      // to the consumer scan — no consumer contract should appear.
+      writeFile('.gitignore', 'generated/\n');
+      writeFile(
+        'generated/auto.cpp',
+        `#include "remote/should_not_appear.h"
+int auto_main() { return 0; }`,
+      );
+
+      const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+      const consumerIds = contracts.filter((c) => c.role === 'consumer').map((c) => c.contractId);
+
+      expect(consumerIds).not.toContain('include::remote/should_not_appear.h');
+    });
+
+    it('skips a provider header whose size exceeds GITNEXUS_MAX_FILE_SIZE', async () => {
+      const previous = process.env.GITNEXUS_MAX_FILE_SIZE;
+      process.env.GITNEXUS_MAX_FILE_SIZE = '1'; // 1 KB cap
+      try {
+        // 4 KB header — comfortably exceeds the cap.
+        const oversized = '#pragma once\n' + 'x'.repeat(4 * 1024);
+        writeFile('huge/big.h', oversized);
+        writeFile('small/tiny.h', '#pragma once');
+
+        const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+        const providerIds = contracts.filter((c) => c.role === 'provider').map((p) => p.contractId);
+
+        expect(providerIds).toContain('include::small/tiny.h');
+        expect(providerIds).not.toContain('include::huge/big.h');
+      } finally {
+        if (previous === undefined) delete process.env.GITNEXUS_MAX_FILE_SIZE;
+        else process.env.GITNEXUS_MAX_FILE_SIZE = previous;
+      }
+    });
+
+    it('skips parsing #include directives in source files exceeding GITNEXUS_MAX_FILE_SIZE', async () => {
+      const previous = process.env.GITNEXUS_MAX_FILE_SIZE;
+      process.env.GITNEXUS_MAX_FILE_SIZE = '1';
+      try {
+        const oversized =
+          '#include "remote/should_not_appear.h"\n' +
+          '// padding to push the file past 1 KB\n' +
+          'x'.repeat(4 * 1024);
+        writeFile('big/main.cpp', oversized);
+
+        const contracts = await extractor.extract(null, tmpDir, makeRepo(tmpDir));
+        const consumerIds = contracts.filter((c) => c.role === 'consumer').map((c) => c.contractId);
+
+        expect(consumerIds).not.toContain('include::remote/should_not_appear.h');
+      } finally {
+        if (previous === undefined) delete process.env.GITNEXUS_MAX_FILE_SIZE;
+        else process.env.GITNEXUS_MAX_FILE_SIZE = previous;
+      }
+    });
+  });
+});
diff --git a/gitnexus/test/unit/group/sync.test.ts b/gitnexus/test/unit/group/sync.test.ts
index 88bc3af2d..9f14db231 100644
--- a/gitnexus/test/unit/group/sync.test.ts
+++ b/gitnexus/test/unit/group/sync.test.ts
@@ -3,6 +3,7 @@ import * as fs from 'node:fs';
 import * as path from 'node:path';
 import * as os from 'node:os';
 import { syncGroup, stableRepoPoolId } from '../../../src/core/group/sync.js';
+import { cleanupTempDir } from '../../helpers/test-db.js';
 import { _captureLogger } from '../../../src/core/logger.js';
 import type {
   GroupConfig,
@@ -583,6 +584,47 @@ service OrderService {
     }
   });
 
+  it('does not extract include contracts during real sync when includes detection is disabled', async () => {
+    // PR #1156 Codex follow-up: ce-code-review T1 — verifies the gate at
+    // sync.ts:174 honors `detect.includes: false`. Mirrors the existing
+    // thrift-off pattern at sync.test.ts:545.
+    const tmpDir = fs.mkdtempSync(path.join(os.tmpdir(), 'gitnexus-sync-includes-off-'));
+    const storageDir = path.join(tmpDir, '.gitnexus');
+    fs.mkdirSync(path.join(tmpDir, 'src'), { recursive: true });
+    fs.mkdirSync(storageDir, { recursive: true });
+    fs.writeFileSync(path.join(tmpDir, 'src', 'view.h'), '#pragma once\nclass View {};');
+
+    const config = makeConfig({ 'app/cpp-lib': 'cpp-lib-repo' });
+    config.detect.http = false;
+    config.detect.grpc = false;
+    config.detect.thrift = false;
+    config.detect.topics = false;
+    config.detect.includes = false;
+
+    const poolAdapter = await import('../../../src/core/lbug/pool-adapter.js');
+    const initSpy = vi.spyOn(poolAdapter, 'initLbug').mockResolvedValue(undefined);
+    const closeSpy = vi.spyOn(poolAdapter, 'closeLbug').mockResolvedValue(undefined);
+
+    try {
+      const result = await syncGroup(config, {
+        resolveRepoHandle: async (_name, groupPath) => ({
+          id: 'cpp-lib-repo',
+          path: groupPath,
+          repoPath: tmpDir,
+          storagePath: storageDir,
+        }),
+        skipWrite: true,
+      });
+
+      expect(result.missingRepos).toHaveLength(0);
+      expect(result.contracts.filter((c) => c.type === 'include')).toHaveLength(0);
+    } finally {
+      initSpy.mockRestore();
+      closeSpy.mockRestore();
+      await cleanupTempDir(tmpDir);
+    }
+  });
+
   it('dedupes duplicate wildcard cross-links during sync', async () => {
     const config = makeConfig({ 'app/provider': 'provider-repo', 'app/consumer': 'consumer-repo' });
     const provider: StoredContract = {
@@ -689,7 +731,12 @@ service OrderService {
       expect(registry.version).toBe(1);
       expect(registry.contracts).toHaveLength(0);
     } finally {
-      fs.rmSync(tmpDir, { recursive: true, force: true });
+      // syncGroup now writes bridge.lbug + WAL/shadow sidecars when
+      // skipWrite is false. On Windows, LadybugDB's checkpoint thread can
+      // briefly outlive closeBridgeDb, holding a Win32 lock on the file.
+      // cleanupTempDir tolerates the documented Windows-native lock codes
+      // (EBUSY/EPERM/EACCES/ENOTEMPTY) with bounded retries.
+      await cleanupTempDir(tmpDir);
     }
   });
 
diff --git a/gitnexus/test/unit/hf-env.test.ts b/gitnexus/test/unit/hf-env.test.ts
index 6a5697059..fa8f23bd3 100644
--- a/gitnexus/test/unit/hf-env.test.ts
+++ b/gitnexus/test/unit/hf-env.test.ts
@@ -1,7 +1,20 @@
-import { describe, it, expect, beforeEach, afterEach } from 'vitest';
+import { describe, it, expect, beforeEach, afterEach, vi } from 'vitest';
 import os from 'node:os';
 import { join } from 'node:path';
-import { applyHfEnvOverrides, type HfEnvSubset } from '../../src/core/embeddings/hf-env.js';
+import {
+  applyHfEnvOverrides,
+  isNetworkFetchError,
+  isHfDownloadFailure,
+  isHfCircuitOpenError,
+  HfDownloadCircuitBreaker,
+  withDownloadTimeout,
+  withHfDownloadRetry,
+  CIRCUIT_OPEN_TAG,
+  HF_MAX_ATTEMPTS,
+  HF_MAX_TIMEOUT_MS,
+  HF_MAX_ATTEMPTS_CAP,
+  type HfEnvSubset,
+} from '../../src/core/embeddings/hf-env.js';
 
 describe('applyHfEnvOverrides', () => {
   let envStub: HfEnvSubset;
@@ -82,3 +95,399 @@ describe('applyHfEnvOverrides', () => {
     expect(envStub.remoteHost).toBe('https://hf-mirror.com/');
   });
 });
+
+describe('isNetworkFetchError', () => {
+  it('returns true for "fetch failed" (the undici error seen on macOS/Node 24)', () => {
+    expect(isNetworkFetchError('fetch failed')).toBe(true);
+  });
+
+  it('returns true for ECONNREFUSED', () => {
+    expect(isNetworkFetchError('connect ECONNREFUSED 13.45.67.89:443')).toBe(true);
+  });
+
+  it('returns true for ENOTFOUND (DNS failure)', () => {
+    expect(isNetworkFetchError('getaddrinfo ENOTFOUND huggingface.co')).toBe(true);
+  });
+
+  it('returns true for ETIMEDOUT', () => {
+    expect(isNetworkFetchError('connect ETIMEDOUT 13.45.67.89:443')).toBe(true);
+  });
+
+  it('returns true for ECONNRESET', () => {
+    expect(isNetworkFetchError('read ECONNRESET')).toBe(true);
+  });
+
+  it('returns false for generic model-load errors (ONNX device failure)', () => {
+    expect(isNetworkFetchError('Failed to initialize CUDA backend')).toBe(false);
+  });
+
+  it('returns false for empty string', () => {
+    expect(isNetworkFetchError('')).toBe(false);
+  });
+
+  it('returns false for module-not-found errors', () => {
+    expect(isNetworkFetchError('Cannot find module onnxruntime-node')).toBe(false);
+  });
+});
+
+describe('isHfCircuitOpenError', () => {
+  it('returns true for a circuit-open tag message', () => {
+    expect(isHfCircuitOpenError(`${CIRCUIT_OPEN_TAG}: circuit is open`)).toBe(true);
+  });
+
+  it('returns false for a plain network error', () => {
+    expect(isHfCircuitOpenError('fetch failed')).toBe(false);
+  });
+});
+
+describe('isHfDownloadFailure', () => {
+  it('returns true for network fetch errors', () => {
+    expect(isHfDownloadFailure('ECONNREFUSED 127.0.0.1:443')).toBe(true);
+  });
+
+  it('returns true for circuit-open errors', () => {
+    expect(isHfDownloadFailure(`${CIRCUIT_OPEN_TAG}: open`)).toBe(true);
+  });
+
+  it('returns false for ONNX device errors', () => {
+    expect(isHfDownloadFailure('Failed to initialize CUDA')).toBe(false);
+  });
+});
+
+describe('HfDownloadCircuitBreaker', () => {
+  it('starts in closed state', () => {
+    const cb = new HfDownloadCircuitBreaker();
+    expect(cb.isOpen()).toBe(false);
+    expect(cb.state).toBe('closed');
+  });
+
+  it('opens after reaching the failure threshold', () => {
+    const cb = new HfDownloadCircuitBreaker(3);
+    cb.recordFailure();
+    cb.recordFailure();
+    expect(cb.isOpen()).toBe(false);
+    cb.recordFailure(); // threshold reached
+    expect(cb.isOpen()).toBe(true);
+    expect(cb.state).toBe('open');
+  });
+
+  it('closes on recordSuccess after being open', () => {
+    const cb = new HfDownloadCircuitBreaker(1);
+    cb.recordFailure();
+    expect(cb.isOpen()).toBe(true);
+    cb.recordSuccess();
+    expect(cb.isOpen()).toBe(false);
+    expect(cb.state).toBe('closed');
+  });
+
+  it('transitions to half-open after the reset timeout', () => {
+    vi.useFakeTimers();
+    try {
+      const cb = new HfDownloadCircuitBreaker(1, 100 /* 100ms */);
+      cb.recordFailure();
+      expect(cb.isOpen()).toBe(true);
+      vi.advanceTimersByTime(200);
+      expect(cb.isOpen()).toBe(false);
+      expect(cb.state).toBe('half-open');
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+
+  it('reset() restores closed state', () => {
+    const cb = new HfDownloadCircuitBreaker(1);
+    cb.recordFailure();
+    expect(cb.isOpen()).toBe(true);
+    cb.reset();
+    expect(cb.isOpen()).toBe(false);
+    expect(cb.state).toBe('closed');
+  });
+
+  it('re-opens when a failure is recorded in half-open state', () => {
+    vi.useFakeTimers();
+    try {
+      const cb = new HfDownloadCircuitBreaker(1, 100 /* 100ms */);
+      cb.recordFailure(); // opens the circuit
+      vi.advanceTimersByTime(200); // advance past reset timeout
+      expect(cb.state).toBe('half-open'); // getter transitions _state to half-open
+      cb.recordFailure(); // failure in half-open → re-opens
+      expect(cb.isOpen()).toBe(true);
+      expect(cb.state).toBe('open');
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+
+  it('closes the circuit when success is recorded in half-open state', () => {
+    vi.useFakeTimers();
+    try {
+      const cb = new HfDownloadCircuitBreaker(1, 100 /* 100ms */);
+      cb.recordFailure(); // opens the circuit
+      vi.advanceTimersByTime(200); // advance past reset timeout
+      expect(cb.state).toBe('half-open');
+      cb.recordSuccess(); // success in half-open → closes
+      expect(cb.isOpen()).toBe(false);
+      expect(cb.state).toBe('closed');
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+});
+
+describe('withDownloadTimeout', () => {
+  it('resolves when fn completes before the timeout', async () => {
+    const result = await withDownloadTimeout(() => Promise.resolve(42), 1_000);
+    expect(result).toBe(42);
+  });
+
+  it('rejects with ETIMEDOUT when fn takes too long', async () => {
+    vi.useFakeTimers();
+    try {
+      const neverResolves = () => new Promise<never>(() => {});
+      const promise = withDownloadTimeout(neverResolves, 20);
+      vi.advanceTimersByTime(30);
+      await expect(promise).rejects.toThrow('ETIMEDOUT');
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+
+  it('propagates non-timeout errors from fn', async () => {
+    await expect(
+      withDownloadTimeout(() => Promise.reject(new Error('download error')), 1_000),
+    ).rejects.toThrow('download error');
+  });
+});
+
+describe('withHfDownloadRetry', () => {
+  it('returns the result on first success', async () => {
+    const fn = vi.fn().mockResolvedValue('ok');
+    const cb = new HfDownloadCircuitBreaker();
+    const result = await withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 });
+    expect(result).toBe('ok');
+    expect(fn).toHaveBeenCalledTimes(1);
+  });
+
+  it('retries on network errors and succeeds on second attempt', async () => {
+    const fn = vi.fn().mockRejectedValueOnce(new Error('fetch failed')).mockResolvedValue('ok');
+    const cb = new HfDownloadCircuitBreaker();
+    const result = await withHfDownloadRetry(fn, {
+      circuit: cb,
+      maxAttempts: 3,
+      baseDelayMs: 0,
+    });
+    expect(result).toBe('ok');
+    expect(fn).toHaveBeenCalledTimes(2);
+  });
+
+  it('throws the last network error after all attempts are exhausted', async () => {
+    const fn = vi.fn().mockRejectedValue(new Error('ECONNREFUSED 127.0.0.1:443'));
+    const cb = new HfDownloadCircuitBreaker(99 /* high threshold */);
+    await expect(
+      withHfDownloadRetry(fn, { circuit: cb, maxAttempts: 3, baseDelayMs: 0 }),
+    ).rejects.toThrow('ECONNREFUSED');
+    expect(fn).toHaveBeenCalledTimes(3);
+  });
+
+  it('does not retry non-network errors', async () => {
+    const fn = vi.fn().mockRejectedValue(new Error('Failed to initialize CUDA backend'));
+    const cb = new HfDownloadCircuitBreaker();
+    await expect(
+      withHfDownloadRetry(fn, { circuit: cb, maxAttempts: 3, baseDelayMs: 0 }),
+    ).rejects.toThrow('Failed to initialize CUDA backend');
+    expect(fn).toHaveBeenCalledTimes(1);
+  });
+
+  it('fails immediately when the circuit is already open', async () => {
+    const fn = vi.fn().mockResolvedValue('ok');
+    const cb = new HfDownloadCircuitBreaker(1);
+    cb.recordFailure(); // open the circuit
+    await expect(withHfDownloadRetry(fn, { circuit: cb })).rejects.toThrow(CIRCUIT_OPEN_TAG);
+    expect(fn).not.toHaveBeenCalled();
+  });
+
+  it('opens the circuit after failureThreshold failures and throws a circuit-open error', async () => {
+    const fn = vi.fn().mockRejectedValue(new Error('ENOTFOUND huggingface.co'));
+    const cb = new HfDownloadCircuitBreaker(2 /* threshold */, 60_000);
+    // First call: 2 attempts, threshold=2 → circuit opens on 2nd failure
+    await expect(
+      withHfDownloadRetry(fn, { circuit: cb, maxAttempts: 2, baseDelayMs: 0 }),
+    ).rejects.toThrow(CIRCUIT_OPEN_TAG);
+    expect(cb.isOpen()).toBe(true);
+  });
+
+  it('calls onRetry with correct arguments on each retry', async () => {
+    const fn = vi
+      .fn()
+      .mockRejectedValueOnce(new Error('fetch failed'))
+      .mockRejectedValueOnce(new Error('fetch failed'))
+      .mockResolvedValue('ok');
+    const cb = new HfDownloadCircuitBreaker(99);
+    const onRetry = vi.fn();
+    await withHfDownloadRetry(fn, { circuit: cb, maxAttempts: 3, baseDelayMs: 0, onRetry });
+    expect(onRetry).toHaveBeenCalledTimes(2);
+    expect(onRetry).toHaveBeenNthCalledWith(
+      1,
+      1,
+      3,
+      expect.objectContaining({ message: 'fetch failed' }),
+    );
+    expect(onRetry).toHaveBeenNthCalledWith(
+      2,
+      2,
+      3,
+      expect.objectContaining({ message: 'fetch failed' }),
+    );
+  });
+
+  it('resets the circuit on success', async () => {
+    const fn = vi.fn().mockResolvedValue('value');
+    const cb = new HfDownloadCircuitBreaker(5);
+    cb.recordFailure();
+    cb.recordFailure(); // 2 failures, circuit still closed
+    await withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 });
+    expect(cb.state).toBe('closed');
+  });
+});
+
+describe('withHfDownloadRetry env overrides', () => {
+  let originalTimeout: string | undefined;
+  let originalMaxAttempts: string | undefined;
+
+  beforeEach(() => {
+    originalTimeout = process.env.HF_DOWNLOAD_TIMEOUT_MS;
+    originalMaxAttempts = process.env.HF_MAX_ATTEMPTS;
+    delete process.env.HF_DOWNLOAD_TIMEOUT_MS;
+    delete process.env.HF_MAX_ATTEMPTS;
+  });
+
+  afterEach(() => {
+    if (originalTimeout === undefined) delete process.env.HF_DOWNLOAD_TIMEOUT_MS;
+    else process.env.HF_DOWNLOAD_TIMEOUT_MS = originalTimeout;
+    if (originalMaxAttempts === undefined) delete process.env.HF_MAX_ATTEMPTS;
+    else process.env.HF_MAX_ATTEMPTS = originalMaxAttempts;
+  });
+
+  it('HF_MAX_ATTEMPTS=1 gives exactly 1 attempt', async () => {
+    process.env.HF_MAX_ATTEMPTS = '1';
+    const fn = vi.fn().mockRejectedValue(new Error('ECONNREFUSED 127.0.0.1:443'));
+    const cb = new HfDownloadCircuitBreaker(99_999 /* high threshold */);
+    await expect(withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 })).rejects.toThrow(
+      'ECONNREFUSED',
+    );
+    expect(fn).toHaveBeenCalledTimes(1);
+  });
+
+  it('HF_MAX_ATTEMPTS=2 gives exactly 2 attempts', async () => {
+    process.env.HF_MAX_ATTEMPTS = '2';
+    const fn = vi.fn().mockRejectedValue(new Error('ENOTFOUND huggingface.co'));
+    const cb = new HfDownloadCircuitBreaker(99_999);
+    await expect(withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 })).rejects.toThrow(
+      'ENOTFOUND',
+    );
+    expect(fn).toHaveBeenCalledTimes(2);
+  });
+
+  it('HF_MAX_ATTEMPTS=abc falls back to the built-in default', async () => {
+    process.env.HF_MAX_ATTEMPTS = 'abc';
+    const fn = vi.fn().mockRejectedValue(new Error('fetch failed'));
+    const cb = new HfDownloadCircuitBreaker(99_999);
+    await expect(withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 })).rejects.toThrow(
+      'fetch failed',
+    );
+    expect(fn).toHaveBeenCalledTimes(HF_MAX_ATTEMPTS);
+  });
+
+  it('HF_MAX_ATTEMPTS=0 falls back to the built-in default', async () => {
+    process.env.HF_MAX_ATTEMPTS = '0';
+    const fn = vi.fn().mockRejectedValue(new Error('fetch failed'));
+    const cb = new HfDownloadCircuitBreaker(99_999);
+    await expect(withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 })).rejects.toThrow(
+      'fetch failed',
+    );
+    expect(fn).toHaveBeenCalledTimes(HF_MAX_ATTEMPTS);
+  });
+
+  it('HF_MAX_ATTEMPTS=-1 falls back to the built-in default', async () => {
+    process.env.HF_MAX_ATTEMPTS = '-1';
+    const fn = vi.fn().mockRejectedValue(new Error('fetch failed'));
+    const cb = new HfDownloadCircuitBreaker(99_999);
+    await expect(withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 })).rejects.toThrow(
+      'fetch failed',
+    );
+    expect(fn).toHaveBeenCalledTimes(HF_MAX_ATTEMPTS);
+  });
+
+  it('HF_MAX_ATTEMPTS is clamped to HF_MAX_ATTEMPTS_CAP', async () => {
+    process.env.HF_MAX_ATTEMPTS = '9999';
+    const fn = vi.fn().mockRejectedValue(new Error('fetch failed'));
+    const cb = new HfDownloadCircuitBreaker(99_999 /* very high threshold */);
+    await expect(withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 })).rejects.toThrow(
+      'fetch failed',
+    );
+    expect(fn).toHaveBeenCalledTimes(HF_MAX_ATTEMPTS_CAP);
+  });
+
+  it('HF_MAX_ATTEMPTS=2.9 is floored to 2', async () => {
+    process.env.HF_MAX_ATTEMPTS = '2.9';
+    const fn = vi.fn().mockRejectedValue(new Error('fetch failed'));
+    const cb = new HfDownloadCircuitBreaker(99_999);
+    await expect(withHfDownloadRetry(fn, { circuit: cb, baseDelayMs: 0 })).rejects.toThrow(
+      'fetch failed',
+    );
+    expect(fn).toHaveBeenCalledTimes(2);
+  });
+
+  it('HF_DOWNLOAD_TIMEOUT_MS is used as the per-attempt timeout when valid', async () => {
+    vi.useFakeTimers();
+    try {
+      process.env.HF_DOWNLOAD_TIMEOUT_MS = '50';
+      const neverResolves = () => new Promise<never>(() => {});
+      const cb = new HfDownloadCircuitBreaker(99);
+      const promise = withHfDownloadRetry(neverResolves, { circuit: cb, maxAttempts: 1 });
+      vi.advanceTimersByTime(100);
+      await expect(promise).rejects.toThrow('ETIMEDOUT');
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+
+  it('HF_DOWNLOAD_TIMEOUT_MS=-1 falls back to the built-in default', async () => {
+    process.env.HF_DOWNLOAD_TIMEOUT_MS = '-1';
+    // Passing explicit timeoutMs=0 (no real wait) so the test doesn't block;
+    // we just verify that the env var rejection causes options.timeoutMs to be
+    // the default constant (not -1) by confirming the resolved value is used.
+    const fn = vi.fn().mockResolvedValue('ok');
+    const cb = new HfDownloadCircuitBreaker(99);
+    // Provide explicit timeoutMs to avoid the default 5-minute wait
+    const result = await withHfDownloadRetry(fn, { circuit: cb, timeoutMs: 100 });
+    expect(result).toBe('ok');
+  });
+
+  it('HF_DOWNLOAD_TIMEOUT_MS is clamped to HF_MAX_TIMEOUT_MS', async () => {
+    vi.useFakeTimers();
+    try {
+      // Set an env value exceeding the 30-minute cap
+      process.env.HF_DOWNLOAD_TIMEOUT_MS = String(HF_MAX_TIMEOUT_MS + 60_000);
+      const neverResolves = () => new Promise<never>(() => {});
+      const cb = new HfDownloadCircuitBreaker(99);
+      const promise = withHfDownloadRetry(neverResolves, { circuit: cb, maxAttempts: 1 });
+      // Advance just past the 30-minute cap
+      vi.advanceTimersByTime(HF_MAX_TIMEOUT_MS + 1);
+      await expect(promise).rejects.toThrow('ETIMEDOUT');
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+
+  it('explicit options override env vars', async () => {
+    process.env.HF_MAX_ATTEMPTS = '5';
+    const fn = vi.fn().mockRejectedValue(new Error('fetch failed'));
+    const cb = new HfDownloadCircuitBreaker(99);
+    // explicit maxAttempts: 2 must win over HF_MAX_ATTEMPTS=5
+    await expect(
+      withHfDownloadRetry(fn, { circuit: cb, maxAttempts: 2, baseDelayMs: 0 }),
+    ).rejects.toThrow('fetch failed');
+    expect(fn).toHaveBeenCalledTimes(2);
+  });
+});
diff --git a/gitnexus/test/unit/ignore-service.test.ts b/gitnexus/test/unit/ignore-service.test.ts
index b4e5cdca1..1e1908137 100644
--- a/gitnexus/test/unit/ignore-service.test.ts
+++ b/gitnexus/test/unit/ignore-service.test.ts
@@ -28,6 +28,8 @@ describe('shouldIgnorePath', () => {
     it.each([
       'node_modules',
       'vendor',
+      'third_party',
+      '3rdparty',
       'venv',
       '.venv',
       '__pycache__',
diff --git a/gitnexus/test/unit/lbug-config-wal.test.ts b/gitnexus/test/unit/lbug-config-wal.test.ts
new file mode 100644
index 000000000..6baea3621
--- /dev/null
+++ b/gitnexus/test/unit/lbug-config-wal.test.ts
@@ -0,0 +1,59 @@
+import { describe, expect, it, vi } from 'vitest';
+import { createLbugDatabase, isWalCorruptionError } from '../../src/core/lbug/lbug-config.js';
+
+describe('isWalCorruptionError', () => {
+  it.each([
+    [
+      'Corrupted wal file',
+      'Runtime exception: Corrupted wal file. Read out invalid WAL record type.',
+    ],
+    ['invalid WAL record', 'Error: invalid WAL record type'],
+    ['WAL checksum', 'Checksum verification failed, the WAL file is corrupted.'],
+    ['WAL + corrupt', 'the WAL file is corrupted'],
+  ])('matches WAL corruption: %s', (_label, msg) => {
+    expect(isWalCorruptionError(msg)).toBe(true);
+    expect(isWalCorruptionError(new Error(msg))).toBe(true);
+  });
+
+  it.each([
+    ['lock error', 'Could not set lock on file : /path/to/db'],
+    ['generic', 'Query failed'],
+    ['not found', 'LadybugDB not found at /path'],
+    ['checksum without WAL', 'Checksum verification failed for parquet file'],
+    ['permission path with WAL', "EACCES: permission denied '/path/to/wal'"],
+    ['schema mismatch WAL', 'schema version mismatch in WAL'],
+  ])('does not match non-WAL error: %s', (_label, msg) => {
+    expect(isWalCorruptionError(msg)).toBe(false);
+  });
+
+  it('handles non-string input', () => {
+    expect(isWalCorruptionError(undefined)).toBe(false);
+    expect(isWalCorruptionError(null)).toBe(false);
+    expect(isWalCorruptionError(42)).toBe(false);
+    expect(isWalCorruptionError(new Error('ok'))).toBe(false);
+  });
+});
+
+describe('createLbugDatabase WAL replay option', () => {
+  it('passes throwOnWalReplayFailure and checksum constructor args explicitly', () => {
+    const Database = vi.fn(function (this: any) {});
+    const lbugModule = { Database } as any;
+
+    createLbugDatabase(lbugModule, '/tmp/lbug', {
+      readOnly: true,
+      throwOnWalReplayFailure: false,
+    });
+
+    expect(Database).toHaveBeenCalledWith(
+      '/tmp/lbug',
+      0,
+      false,
+      true,
+      expect.any(Number),
+      true,
+      -1,
+      false,
+      true,
+    );
+  });
+});
diff --git a/gitnexus/test/unit/mcp-wal-feedback.test.ts b/gitnexus/test/unit/mcp-wal-feedback.test.ts
new file mode 100644
index 000000000..e387e0b97
--- /dev/null
+++ b/gitnexus/test/unit/mcp-wal-feedback.test.ts
@@ -0,0 +1,160 @@
+/**
+ * Tests for WAL corruption feedback in MCP error responses (#1402).
+ */
+import { beforeEach, describe, expect, it, vi } from 'vitest';
+
+const { lbugMocks, platformMocks, repoMocks } = vi.hoisted(() => ({
+  lbugMocks: {
+    initLbug: vi.fn().mockResolvedValue(undefined),
+    executeQuery: vi.fn(),
+    executeParameterized: vi.fn(),
+    closeLbug: vi.fn().mockResolvedValue(undefined),
+    isLbugReady: vi.fn().mockReturnValue(true),
+    isWriteQuery: vi.fn().mockReturnValue(false),
+  },
+  platformMocks: {
+    isVectorExtensionSupportedByPlatform: vi.fn().mockReturnValue(true),
+  },
+  repoMocks: {
+    listRegisteredRepos: vi.fn(),
+  },
+}));
+
+vi.mock('../../src/core/lbug/pool-adapter.js', async (importOriginal) => {
+  const actual = await importOriginal();
+  return { ...actual, ...lbugMocks };
+});
+
+vi.mock('../../src/mcp/core/lbug-adapter.js', async (importOriginal) => {
+  const actual = await importOriginal();
+  return { ...actual, ...lbugMocks };
+});
+
+vi.mock('../../src/storage/repo-manager.js', () => ({
+  listRegisteredRepos: repoMocks.listRegisteredRepos,
+  cleanupOldKuzuFiles: vi.fn().mockResolvedValue({ found: false, needsReindex: false }),
+  findSiblingClones: vi.fn().mockResolvedValue([]),
+}));
+
+vi.mock('../../src/core/git-staleness.js', () => ({
+  checkStaleness: vi.fn().mockReturnValue({ isStale: false, commitsBehind: 0 }),
+  checkCwdMatch: vi.fn().mockResolvedValue({ match: 'none' }),
+}));
+
+vi.mock('../../src/core/platform/capabilities.js', async (importOriginal) => {
+  const actual = await importOriginal<typeof import('../../src/core/platform/capabilities.js')>();
+  return {
+    ...actual,
+    isVectorExtensionSupportedByPlatform: platformMocks.isVectorExtensionSupportedByPlatform,
+  };
+});
+
+vi.mock('../../src/core/search/bm25-index.js', () => ({
+  searchFTSFromLbug: vi.fn().mockResolvedValue([]),
+}));
+
+vi.mock('../../src/mcp/core/embedder.js', () => ({
+  embedQuery: vi.fn().mockResolvedValue([]),
+  getEmbeddingDims: vi.fn().mockReturnValue(384),
+}));
+
+import { LocalBackend } from '../../src/mcp/local/local-backend.js';
+
+const MOCK_REPO_ENTRY = {
+  name: 'test-repo',
+  path: '/tmp/test',
+  storagePath: '/tmp/test/.gitnexus',
+  indexedAt: '2026-05-01T00:00:00Z',
+  lastCommit: 'abc1234',
+};
+
+async function makeBackend(): Promise<LocalBackend> {
+  const backend = new LocalBackend();
+  await backend.init();
+  return backend;
+}
+
+describe('WAL corruption feedback in MCP responses (#1402)', () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+    lbugMocks.initLbug.mockResolvedValue(undefined);
+    lbugMocks.executeQuery.mockResolvedValue([]);
+    lbugMocks.executeParameterized.mockResolvedValue([]);
+    lbugMocks.isLbugReady.mockReturnValue(true);
+    lbugMocks.isWriteQuery.mockReturnValue(false);
+    repoMocks.listRegisteredRepos.mockResolvedValue([MOCK_REPO_ENTRY]);
+  });
+
+  it('impact returns WAL suggestion on corrupted WAL error', async () => {
+    const backend = await makeBackend();
+    lbugMocks.executeParameterized.mockRejectedValueOnce(
+      new Error('Runtime exception: Corrupted wal file. Read out invalid WAL record type.'),
+    );
+
+    const result = await backend.callTool('impact', {
+      repo: 'test-repo',
+      target: 'MyClass',
+      direction: 'upstream',
+    });
+
+    expect(result.error).toBeDefined();
+    expect(result.suggestion).toBe(
+      'The graph query failed — try gitnexus context <symbol> as a fallback',
+    );
+    expect(result.recoverySuggestion).toBeDefined();
+  });
+
+  it('cypher returns WAL recoverySuggestion on corrupted WAL error', async () => {
+    const backend = await makeBackend();
+    lbugMocks.executeQuery.mockRejectedValueOnce(new Error('Corrupted wal file'));
+
+    const result = await backend.callTool('cypher', {
+      repo: 'test-repo',
+      query: 'MATCH (n) RETURN n LIMIT 1',
+    });
+
+    expect(result.error).toBe('Corrupted wal file');
+    expect(result.recoverySuggestion).toBeDefined();
+  });
+
+  it('context returns WAL recoverySuggestion on corrupted WAL error', async () => {
+    const backend = await makeBackend();
+    lbugMocks.executeParameterized.mockRejectedValueOnce(new Error('Corrupted wal file'));
+
+    const result = await backend.callTool('context', {
+      repo: 'test-repo',
+      name: 'MyClass',
+    });
+
+    expect(result.error).toBe('Corrupted wal file');
+    expect(result.recoverySuggestion).toBeDefined();
+  });
+
+  it('non-WAL errors do not include WAL suggestion', async () => {
+    const backend = await makeBackend();
+    lbugMocks.executeParameterized.mockRejectedValueOnce(new Error('Some other error'));
+
+    const result = await backend.callTool('impact', {
+      repo: 'test-repo',
+      target: 'MyClass',
+      direction: 'upstream',
+    });
+
+    expect(result.error).toBeDefined();
+    expect(result.suggestion).toBe(
+      'The graph query failed — try gitnexus context <symbol> as a fallback',
+    );
+  });
+
+  it('context preserves non-WAL throw behavior', async () => {
+    const backend = await makeBackend();
+    lbugMocks.executeParameterized.mockRejectedValueOnce(new Error('Some other error'));
+
+    await expect(
+      backend.callTool('context', {
+        repo: 'test-repo',
+        name: 'MyClass',
+      }),
+    ).rejects.toThrow('Some other error');
+  });
+});
diff --git a/gitnexus/test/unit/mcp/group-repo-routing.test.ts b/gitnexus/test/unit/mcp/group-repo-routing.test.ts
index ddf7d03a5..474abeaa8 100644
--- a/gitnexus/test/unit/mcp/group-repo-routing.test.ts
+++ b/gitnexus/test/unit/mcp/group-repo-routing.test.ts
@@ -32,7 +32,7 @@ vi.mock('../../../src/storage/repo-manager.js', () => ({
 }));
 
 vi.mock('../../../src/core/search/bm25-index.js', () => ({
-  searchFTSFromLbug: vi.fn().mockResolvedValue([]),
+  searchFTSFromLbug: vi.fn().mockResolvedValue({ results: [], ftsAvailable: true }),
 }));
 
 vi.mock('../../../src/mcp/core/embedder.js', () => ({
diff --git a/gitnexus/test/unit/pool-wal-recovery.test.ts b/gitnexus/test/unit/pool-wal-recovery.test.ts
new file mode 100644
index 000000000..19b24c583
--- /dev/null
+++ b/gitnexus/test/unit/pool-wal-recovery.test.ts
@@ -0,0 +1,177 @@
+/**
+ * Tests for WAL corruption recovery in the connection pool (#1402).
+ *
+ * Mocks createLbugDatabase and fs to verify quarantine + retry behavior
+ * without needing a real LadybugDB instance or corrupted WAL file.
+ */
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+const { stderrWriteMock } = vi.hoisted(() => ({
+  stderrWriteMock: vi.fn(),
+}));
+
+vi.mock('fs/promises', () => ({
+  default: {
+    stat: vi.fn().mockResolvedValue({}),
+    unlink: vi.fn().mockResolvedValue(undefined),
+    rename: vi.fn().mockResolvedValue(undefined),
+  },
+}));
+
+vi.mock('@ladybugdb/core', () => ({
+  default: {
+    Database: vi.fn(),
+    Connection: vi.fn(function (this: any) {
+      this.close = vi.fn().mockResolvedValue(undefined);
+    }),
+  },
+}));
+
+vi.mock('../../src/core/lbug/lbug-adapter.js', () => ({
+  loadFTSExtension: vi.fn().mockResolvedValue(true),
+}));
+
+vi.mock('../../src/core/lbug/lbug-config.js', () => ({
+  createLbugDatabase: vi.fn(),
+  LBUG_MAX_DB_SIZE: 1024,
+  isWalCorruptionError: vi.fn((err: unknown) => {
+    const msg = err instanceof Error ? err.message : String(err ?? '');
+    return /corrupt(ed)?\s+wal|invalid\s+wal\s+record/i.test(msg);
+  }),
+}));
+
+vi.mock('../../src/mcp/stdio-capture.js', () => ({
+  realStdoutWrite: vi.fn(),
+  realStderrWrite: stderrWriteMock,
+  setActiveStdoutWrite: vi.fn(),
+  getActiveStdoutWrite: vi.fn(() => vi.fn()),
+}));
+
+import fs from 'fs/promises';
+import { createLbugDatabase } from '../../src/core/lbug/lbug-config.js';
+
+const { closeLbug } = await import('../../src/core/lbug/pool-adapter.js');
+
+const mockInit = vi.fn().mockResolvedValue(undefined);
+const mockClose = vi.fn().mockResolvedValue(undefined);
+
+function makeMockDb() {
+  return { init: mockInit, close: mockClose, _isClosed: false } as any;
+}
+
+describe('WAL corruption recovery in doInitLbug (#1402)', () => {
+  beforeEach(() => {
+    (createLbugDatabase as any).mockReset();
+    (fs.stat as any).mockReset();
+    (fs.rename as any).mockReset();
+    mockInit.mockReset();
+    mockClose.mockReset();
+    mockInit.mockResolvedValue(undefined);
+    mockClose.mockResolvedValue(undefined);
+    (fs.stat as any).mockResolvedValue({});
+    (fs.rename as any).mockResolvedValue(undefined);
+  });
+
+  afterEach(async () => {
+    vi.useRealTimers();
+    await closeLbug().catch(() => {});
+    vi.clearAllMocks();
+  });
+
+  it('retries with WAL quarantine on corrupted WAL init error', async () => {
+    const { initLbug } = await import('../../src/core/lbug/pool-adapter.js');
+    const dbPath = '/tmp/test-wal-recovery/lbug';
+
+    const badDb = makeMockDb();
+    const goodDb = makeMockDb();
+    badDb.init = vi.fn().mockRejectedValueOnce(new Error('Corrupted wal file'));
+    (createLbugDatabase as any).mockReturnValueOnce(badDb).mockReturnValueOnce(goodDb);
+
+    await initLbug('test-repo-init', dbPath);
+
+    expect(badDb.init).toHaveBeenCalledTimes(1);
+    expect(createLbugDatabase).toHaveBeenCalledTimes(2);
+    expect(createLbugDatabase).toHaveBeenCalledWith(
+      expect.anything(),
+      dbPath,
+      expect.objectContaining({
+        readOnly: true,
+        throwOnWalReplayFailure: false,
+      }),
+    );
+    expect(fs.rename).toHaveBeenCalledWith(
+      dbPath + '.wal',
+      expect.stringContaining('.wal.corrupt.'),
+    );
+    expect(stderrWriteMock).toHaveBeenCalledWith(
+      expect.stringContaining('WAL quarantined for test-repo-init'),
+    );
+  });
+
+  it('does not quarantine on lock error (preserves existing lock retry)', async () => {
+    const { initLbug } = await import('../../src/core/lbug/pool-adapter.js');
+    const setTimeoutSpy = vi.spyOn(global, 'setTimeout').mockImplementation((callback: any) => {
+      callback();
+      return 0 as any;
+    });
+    const dbPath = '/tmp/test-wal-recovery/lbug';
+
+    (createLbugDatabase as any).mockImplementation(() => {
+      throw new Error('Could not set lock on file');
+    });
+
+    try {
+      await expect(initLbug('test-repo-lock', dbPath)).rejects.toThrow();
+    } finally {
+      setTimeoutSpy.mockRestore();
+    }
+
+    expect(fs.rename).not.toHaveBeenCalled();
+  });
+
+  it('throws with analyze suggestion after retry also fails', async () => {
+    const { initLbug } = await import('../../src/core/lbug/pool-adapter.js');
+    const dbPath = '/tmp/test-wal-recovery/lbug';
+
+    (createLbugDatabase as any)
+      .mockImplementationOnce(() => {
+        throw new Error('Corrupted wal file');
+      })
+      .mockImplementationOnce(() => {
+        throw new Error('Still broken');
+      });
+
+    await expect(initLbug('test-repo-fail', dbPath)).rejects.toThrow(/gitnexus analyze/);
+    expect(createLbugDatabase).toHaveBeenCalledTimes(2);
+  });
+
+  it('does not reuse poisoned state after WAL failure', async () => {
+    const { initLbug, isLbugReady: ready } = await import('../../src/core/lbug/pool-adapter.js');
+    const dbPath = '/tmp/test-wal-recovery/lbug';
+
+    (createLbugDatabase as any)
+      .mockImplementationOnce(() => {
+        throw new Error('Corrupted wal file');
+      })
+      .mockImplementationOnce(() => {
+        throw new Error('Still broken');
+      });
+
+    await expect(initLbug('test-repo-nocache', dbPath)).rejects.toThrow();
+
+    expect(ready('test-repo-nocache')).toBe(false);
+  });
+
+  it('handles quarantine gracefully when .wal file does not exist', async () => {
+    const { initLbug } = await import('../../src/core/lbug/pool-adapter.js');
+    const dbPath = '/tmp/test-wal-recovery/lbug';
+
+    (fs.rename as any).mockRejectedValueOnce(new Error('ENOENT: no such file'));
+
+    (createLbugDatabase as any).mockImplementationOnce(() => {
+      throw new Error('Corrupted wal file');
+    });
+
+    await expect(initLbug('test-repo-enoent', dbPath)).rejects.toThrow(/gitnexus analyze/);
+  });
+});
diff --git a/gitnexus/test/unit/publish.test.ts b/gitnexus/test/unit/publish.test.ts
new file mode 100644
index 000000000..2ba767594
--- /dev/null
+++ b/gitnexus/test/unit/publish.test.ts
@@ -0,0 +1,316 @@
+import { afterEach, beforeEach, describe, expect, it, test, vi } from 'vitest';
+import fs from 'fs/promises';
+import os from 'os';
+import path from 'path';
+import { performance } from 'node:perf_hooks';
+import {
+  buildUqDispatchPayload,
+  isValidOwnerRepo,
+  parseOwnerRepoFromRemote,
+  stripGitSuffix,
+  UNDERSTAND_QUICKLY_TOKEN_ENV,
+} from 'gitnexus-shared';
+
+describe('understand-quickly helpers (gitnexus-shared)', () => {
+  describe('isValidOwnerRepo', () => {
+    it.each([
+      ['looptech-ai/understand-quickly', true],
+      ['abhigyanpatwari/GitNexus', true],
+      // LOW 8: GitHub user/org slugs are alnum/hyphen only — no underscore.
+      ['Some_Org/Some.Repo-2', false],
+      ['', false],
+      ['just-a-name', false],
+      ['/Users/me/code/repo', false],
+      ['org/with spaces', false],
+      ['org//double', false],
+      // LOW 8 additions:
+      ['some_org/repo', false], // underscore in owner — invalid
+      ['-org/repo', false], // leading hyphen — invalid
+      ['org-/repo', false], // trailing hyphen — GitHub rejects at account creation; we mirror that here
+      ['org/repo_with_underscore', true],
+      ['org/.dotfile', true], // repos may start with dot
+    ])('returns %s for %j', (id, expected) => {
+      expect(isValidOwnerRepo(id as string)).toBe(expected);
+    });
+  });
+
+  describe('stripGitSuffix (BLOCKER 1 — ReDoS-safe)', () => {
+    it.each([
+      ['https://github.com/o/r.git', 'https://github.com/o/r'],
+      ['https://github.com/o/r.git/', 'https://github.com/o/r'],
+      ['https://github.com/o/r/', 'https://github.com/o/r'],
+      ['https://github.com/o/r', 'https://github.com/o/r'],
+      ['https://github.com/o/r.GIT', 'https://github.com/o/r'],
+      ['https://github.com/o/r//', 'https://github.com/o/r'],
+      ['', ''],
+      ['/', ''],
+    ])('strips %j -> %j', (input, expected) => {
+      expect(stripGitSuffix(input)).toBe(expected);
+    });
+
+    test('linear time on adversarial trailing slashes (regression for ReDoS)', () => {
+      const adversarial = 'https://github.com/o/r' + '/'.repeat(10_000);
+      const start = performance.now();
+      const result = stripGitSuffix(adversarial);
+      const elapsed = performance.now() - start;
+      expect(result).toBe('https://github.com/o/r');
+      expect(elapsed).toBeLessThan(50); // generous; should be sub-millisecond
+    });
+
+    test('parseOwnerRepoFromRemote terminates quickly on adversarial input', () => {
+      const adversarial = 'https://github.com/o/r.git' + '/'.repeat(10_000);
+      const start = performance.now();
+      const result = parseOwnerRepoFromRemote(adversarial);
+      const elapsed = performance.now() - start;
+      expect(result).toBe('o/r');
+      expect(elapsed).toBeLessThan(50);
+    });
+  });
+
+  describe('parseOwnerRepoFromRemote', () => {
+    it.each([
+      ['git@github.com:looptech-ai/understand-quickly.git', 'looptech-ai/understand-quickly'],
+      ['https://github.com/looptech-ai/understand-quickly', 'looptech-ai/understand-quickly'],
+      ['https://github.com/looptech-ai/understand-quickly.git', 'looptech-ai/understand-quickly'],
+      ['ssh://git@github.com/abhigyanpatwari/GitNexus.git', 'abhigyanpatwari/GitNexus'],
+    ])('parses %s -> %s', (url, expected) => {
+      expect(parseOwnerRepoFromRemote(url)).toBe(expected);
+    });
+
+    // LOW 9: non-GitHub remotes must be rejected — a wrong id is worse
+    // than no id, since the user can always pass --id explicitly.
+    it.each([
+      ['https://gitlab.example.com/group/sub/project.git'],
+      ['git@gitlab.example.com:group/sub/project.git'],
+      ['https://bitbucket.org/team/repo.git'],
+    ])('returns null for non-GitHub host %j', (input) => {
+      expect(parseOwnerRepoFromRemote(input)).toBeNull();
+    });
+
+    it.each([null, undefined, '', '   ', 'not-a-url', 'https://github.com/'])(
+      'returns null for %j',
+      (input) => {
+        expect(parseOwnerRepoFromRemote(input as string | null | undefined)).toBeNull();
+      },
+    );
+  });
+
+  describe('buildUqDispatchPayload', () => {
+    it('wraps the id in the registry-expected event shape', () => {
+      expect(buildUqDispatchPayload('looptech-ai/understand-quickly')).toEqual({
+        event_type: 'sync-entry',
+        client_payload: { id: 'looptech-ai/understand-quickly' },
+      });
+    });
+
+    it('throws on a malformed id rather than building an invalid payload', () => {
+      expect(() => buildUqDispatchPayload('just-a-name')).toThrow(/owner\/repo/);
+      expect(() => buildUqDispatchPayload('/Users/me/repo')).toThrow(/owner\/repo/);
+    });
+  });
+});
+
+describe('publishCommand (no-token no-op)', () => {
+  let tempDir: string;
+  let originalToken: string | undefined;
+  let exitCodeBefore: number | undefined;
+
+  beforeEach(async () => {
+    vi.resetModules();
+    tempDir = await fs.mkdtemp(path.join(os.tmpdir(), 'gn-publish-test-'));
+    // Simulate an existing index so hasIndex() returns true.
+    await fs.mkdir(path.join(tempDir, '.gitnexus'), { recursive: true });
+    await fs.writeFile(
+      path.join(tempDir, '.gitnexus', 'meta.json'),
+      JSON.stringify({ repoPath: tempDir, lastCommit: '', indexedAt: '' }),
+      'utf-8',
+    );
+    originalToken = process.env[UNDERSTAND_QUICKLY_TOKEN_ENV];
+    delete process.env[UNDERSTAND_QUICKLY_TOKEN_ENV];
+    exitCodeBefore = process.exitCode;
+    process.exitCode = 0;
+  });
+
+  afterEach(async () => {
+    if (originalToken !== undefined) {
+      process.env[UNDERSTAND_QUICKLY_TOKEN_ENV] = originalToken;
+    } else {
+      delete process.env[UNDERSTAND_QUICKLY_TOKEN_ENV];
+    }
+    process.exitCode = exitCodeBefore;
+    await fs.rm(tempDir, { recursive: true, force: true });
+  });
+
+  it('exits 0 without firing a network call when the token is unset', async () => {
+    const fetchSpy = vi.spyOn(globalThis, 'fetch').mockImplementation(() => {
+      throw new Error('publishCommand should NOT call fetch when the token is missing');
+    });
+
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, { id: 'looptech-ai/understand-quickly', skipGit: true });
+
+    expect(fetchSpy).not.toHaveBeenCalled();
+    expect(process.exitCode ?? 0).toBe(0);
+    fetchSpy.mockRestore();
+  });
+
+  it('exits 0 with no token even when no index/repo exists (BLOCKER 2)', async () => {
+    // Per the README, CLI --help, and PR body: without a token, the
+    // command must be a no-op even if the repo lacks `.gitnexus/`.
+    const noIndexDir = await fs.mkdtemp(path.join(os.tmpdir(), 'gn-publish-noidx-'));
+    try {
+      const fetchSpy = vi.spyOn(globalThis, 'fetch').mockImplementation(() => {
+        throw new Error('publishCommand should NOT call fetch when the token is missing');
+      });
+      const { publishCommand } = await import('../../src/cli/publish.js');
+      await publishCommand(noIndexDir, {
+        id: 'looptech-ai/understand-quickly',
+        skipGit: true,
+      });
+      expect(fetchSpy).not.toHaveBeenCalled();
+      expect(process.exitCode ?? 0).toBe(0);
+      fetchSpy.mockRestore();
+    } finally {
+      await fs.rm(noIndexDir, { recursive: true, force: true });
+    }
+  });
+});
+
+describe('publishCommand response branches (MEDIUM 5)', () => {
+  let tempDir: string;
+  let originalToken: string | undefined;
+  let exitCodeBefore: number | undefined;
+  let fetchSpy: ReturnType<typeof vi.spyOn>;
+
+  beforeEach(async () => {
+    vi.resetModules();
+    tempDir = await fs.mkdtemp(path.join(os.tmpdir(), 'gn-publish-resp-'));
+    await fs.mkdir(path.join(tempDir, '.gitnexus'), { recursive: true });
+    await fs.writeFile(
+      path.join(tempDir, '.gitnexus', 'meta.json'),
+      JSON.stringify({ repoPath: tempDir, lastCommit: '', indexedAt: '' }),
+      'utf-8',
+    );
+    originalToken = process.env[UNDERSTAND_QUICKLY_TOKEN_ENV];
+    process.env[UNDERSTAND_QUICKLY_TOKEN_ENV] = 'pat_test';
+    exitCodeBefore = process.exitCode;
+    process.exitCode = 0;
+    fetchSpy = vi.spyOn(globalThis, 'fetch');
+  });
+
+  afterEach(async () => {
+    if (originalToken !== undefined) {
+      process.env[UNDERSTAND_QUICKLY_TOKEN_ENV] = originalToken;
+    } else {
+      delete process.env[UNDERSTAND_QUICKLY_TOKEN_ENV];
+    }
+    process.exitCode = exitCodeBefore;
+    vi.restoreAllMocks();
+    await fs.rm(tempDir, { recursive: true, force: true });
+  });
+
+  function mockResponse(status: number, body = '') {
+    fetchSpy.mockResolvedValueOnce({
+      status,
+      ok: status >= 200 && status < 300,
+      text: async () => body,
+      body: { cancel: async () => {} },
+      headers: new Headers(),
+    } as unknown as Response);
+  }
+
+  it('204 → exit 0 with success message', async () => {
+    mockResponse(204);
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, {
+      id: 'looptech-ai/understand-quickly',
+      skipGit: true,
+    });
+    expect(fetchSpy).toHaveBeenCalledTimes(1);
+    expect(process.exitCode ?? 0).toBe(0);
+  });
+
+  it('401 → exit 1 with PAT-invalid hint', async () => {
+    mockResponse(401, '{"message":"Bad credentials"}');
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, {
+      id: 'looptech-ai/understand-quickly',
+      skipGit: true,
+    });
+    expect(process.exitCode).toBe(1);
+  });
+
+  it('403 → exit 1 with scope-missing hint', async () => {
+    mockResponse(403, '{"message":"Resource not accessible"}');
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, {
+      id: 'looptech-ai/understand-quickly',
+      skipGit: true,
+    });
+    expect(process.exitCode).toBe(1);
+  });
+
+  it('404 → exit 1 with repo-access hint', async () => {
+    mockResponse(404, '{"message":"Not Found"}');
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, {
+      id: 'looptech-ai/understand-quickly',
+      skipGit: true,
+    });
+    expect(process.exitCode).toBe(1);
+  });
+
+  it('5xx → exit 1 with raw body', async () => {
+    mockResponse(503, 'gateway timeout');
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, {
+      id: 'looptech-ai/understand-quickly',
+      skipGit: true,
+    });
+    expect(process.exitCode).toBe(1);
+  });
+
+  it('network throw → exit 1', async () => {
+    fetchSpy.mockRejectedValueOnce(new Error('ECONNRESET'));
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, {
+      id: 'looptech-ai/understand-quickly',
+      skipGit: true,
+    });
+    expect(process.exitCode).toBe(1);
+  });
+
+  it('TimeoutError (HIGH 4 — fetch timeout) → exit 1 with timed-out message', async () => {
+    // `AbortSignal.timeout()` throws a real `DOMException` with
+    // `name === 'TimeoutError'`. Faking it as `Error{name:'AbortError'}`
+    // (the previous shape of this test) hid a mismatch in publish.ts —
+    // the catch branch only matched 'AbortError' and the user-facing
+    // "timed out" message never fired in production.
+    const abort = new DOMException('The operation was aborted due to timeout', 'TimeoutError');
+    fetchSpy.mockRejectedValueOnce(abort);
+    const errSpy = vi.spyOn(process.stderr, 'write').mockImplementation(() => true);
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, {
+      id: 'looptech-ai/understand-quickly',
+      skipGit: true,
+    });
+    expect(process.exitCode).toBe(1);
+    const written = errSpy.mock.calls.map((c) => String(c[0])).join('');
+    expect(written).toMatch(/timed out/i);
+    errSpy.mockRestore();
+  });
+
+  it('token never appears in any logged output', async () => {
+    process.env[UNDERSTAND_QUICKLY_TOKEN_ENV] = 'pat_secret_value';
+    mockResponse(401, '');
+    const errSpy = vi.spyOn(process.stderr, 'write').mockImplementation(() => true);
+    const { publishCommand } = await import('../../src/cli/publish.js');
+    await publishCommand(tempDir, {
+      id: 'looptech-ai/understand-quickly',
+      skipGit: true,
+    });
+    const written = errSpy.mock.calls.map((c) => String(c[0])).join('');
+    expect(written).not.toContain('pat_secret_value');
+    errSpy.mockRestore();
+  });
+});
diff --git a/gitnexus/test/unit/schema.test.ts b/gitnexus/test/unit/schema.test.ts
index cab80984a..17db9bdf9 100644
--- a/gitnexus/test/unit/schema.test.ts
+++ b/gitnexus/test/unit/schema.test.ts
@@ -13,6 +13,7 @@ import {
   CLASS_SCHEMA,
   INTERFACE_SCHEMA,
   METHOD_SCHEMA,
+  PROPERTY_SCHEMA,
   CODE_ELEMENT_SCHEMA,
   COMMUNITY_SCHEMA,
   PROCESS_SCHEMA,
@@ -117,6 +118,11 @@ describe('LadybugDB Schema', () => {
       expect(FUNCTION_SCHEMA).toContain('isExported BOOLEAN');
     });
 
+    it('Property schema preserves declaredType', () => {
+      expect(SCHEMA_QUERIES).toContain(PROPERTY_SCHEMA);
+      expect(PROPERTY_SCHEMA).toContain('declaredType STRING');
+    });
+
     it('Community schema has heuristicLabel and cohesion', () => {
       expect(COMMUNITY_SCHEMA).toContain('heuristicLabel STRING');
       expect(COMMUNITY_SCHEMA).toContain('cohesion DOUBLE');
diff --git a/gitnexus/test/unit/scope-resolution/csharp/csharp-captures.test.ts b/gitnexus/test/unit/scope-resolution/csharp/csharp-captures.test.ts
index c8111d203..9fa9bf181 100644
--- a/gitnexus/test/unit/scope-resolution/csharp/csharp-captures.test.ts
+++ b/gitnexus/test/unit/scope-resolution/csharp/csharp-captures.test.ts
@@ -422,4 +422,47 @@ describe('emitCsharpScopeCaptures — references', () => {
     expect(m!['@reference.receiver'].text).toBe('obj');
     expect(m!['@reference.name'].text).toBe('Name');
   });
+
+  it('captures member reads `obj.Name`', () => {
+    const m = findMatch('class A { void M(User obj) { var name = obj.Name; } }', (t) =>
+      t.includes('@reference.read.member'),
+    );
+    expect(m).toBeDefined();
+    expect(m!['@reference.receiver'].text).toBe('obj');
+    expect(m!['@reference.name'].text).toBe('Name');
+  });
+
+  it('does not capture member calls as member reads', () => {
+    const matches = emitCsharpScopeCaptures(
+      'class A { void M(User obj) { obj.Save(); } }',
+      'test.cs',
+    );
+    expect(matches.some((m) => '@reference.call.member' in m)).toBe(true);
+    expect(matches.some((m) => '@reference.read.member' in m)).toBe(false);
+  });
+
+  it('captures generic type arguments as type references', () => {
+    const matches = emitCsharpScopeCaptures(
+      'class A : IEntityTypeConfiguration<USER_INFO> { public Task<List<USER_INFO>> Load(List<USER_INFO> users) => null!; }',
+      'test.cs',
+    );
+    const names = matches
+      .filter((m) => '@reference.type' in m)
+      .map((m) => m['@reference.name'].text);
+
+    expect(names).toContain('USER_INFO');
+    expect(names).not.toContain('string');
+  });
+
+  it('captures call-site generic type arguments as type references', () => {
+    const matches = emitCsharpScopeCaptures(
+      'class A { void M(IRepo repo) { repo.Get<USER_INFO>(); } }',
+      'test.cs',
+    );
+    const names = matches
+      .filter((m) => '@reference.type' in m)
+      .map((m) => m['@reference.name'].text);
+
+    expect(names).toContain('USER_INFO');
+  });
 });
diff --git a/gitnexus/test/unit/staleness.test.ts b/gitnexus/test/unit/staleness.test.ts
index b8ea398ff..1645d9987 100644
--- a/gitnexus/test/unit/staleness.test.ts
+++ b/gitnexus/test/unit/staleness.test.ts
@@ -8,7 +8,7 @@
  */
 import { describe, it, expect } from 'vitest';
 import { execFileSync } from 'child_process';
-import { checkStaleness } from '../../src/core/git-staleness.js';
+import { checkStaleness, checkStalenessAsync } from '../../src/core/git-staleness.js';
 
 // We test checkStaleness with a real git repo (the project itself)
 // since mocking execFileSync across ESM modules is complex.
@@ -65,3 +65,82 @@ describe('checkStaleness', () => {
     expect(result.commitsBehind).toBe(0);
   });
 });
+
+describe('checkStalenessAsync', () => {
+  it('returns not stale when HEAD matches lastCommit', async () => {
+    let headCommit: string;
+    try {
+      headCommit = execFileSync('git', ['rev-parse', 'HEAD'], {
+        encoding: 'utf-8',
+        stdio: ['pipe', 'pipe', 'pipe'],
+      }).trim();
+    } catch {
+      return;
+    }
+
+    const result = await checkStalenessAsync(process.cwd(), headCommit);
+    expect(result.isStale).toBe(false);
+    expect(result.commitsBehind).toBe(0);
+    expect(result.hint).toBeUndefined();
+  });
+
+  it('returns stale when lastCommit is behind HEAD', async () => {
+    let previousCommit: string;
+    try {
+      previousCommit = execFileSync('git', ['rev-parse', 'HEAD~1'], {
+        encoding: 'utf-8',
+        stdio: ['pipe', 'pipe', 'pipe'],
+      }).trim();
+    } catch {
+      return;
+    }
+
+    if (!previousCommit) return;
+
+    const result = await checkStalenessAsync(process.cwd(), previousCommit);
+    expect(result.isStale).toBe(true);
+    expect(result.commitsBehind).toBeGreaterThan(0);
+    expect(result.hint).toContain('behind HEAD');
+  });
+
+  it('fails open when git command fails (e.g., invalid path)', async () => {
+    const result = await checkStalenessAsync('/nonexistent/path', 'abc123');
+    expect(result.isStale).toBe(false);
+    expect(result.commitsBehind).toBe(0);
+  });
+
+  it('fails open with invalid commit hash', async () => {
+    const result = await checkStalenessAsync(process.cwd(), 'not-a-real-commit-hash');
+    expect(result.isStale).toBe(false);
+    expect(result.commitsBehind).toBe(0);
+  });
+
+  it('parallel calls complete faster than sequential', async () => {
+    let headCommit: string;
+    try {
+      headCommit = execFileSync('git', ['rev-parse', 'HEAD'], {
+        encoding: 'utf-8',
+        stdio: ['pipe', 'pipe', 'pipe'],
+      }).trim();
+    } catch {
+      return;
+    }
+
+    const cwd = process.cwd();
+    const N = 10;
+
+    // Parallel
+    const t0 = performance.now();
+    await Promise.all(Array.from({ length: N }, () => checkStalenessAsync(cwd, headCommit)));
+    const parallelMs = performance.now() - t0;
+
+    // Sequential sync
+    const t1 = performance.now();
+    for (let i = 0; i < N; i++) checkStaleness(cwd, headCommit);
+    const sequentialMs = performance.now() - t1;
+
+    // Parallel should be meaningfully faster than sequential.
+    // Use a generous ratio to avoid flakiness on slow CI machines.
+    expect(parallelMs).toBeLessThan(sequentialMs * 1.5);
+  });
+});
diff --git a/gitnexus/test/unit/u8-redos-resource-exhaustion.test.ts b/gitnexus/test/unit/u8-redos-resource-exhaustion.test.ts
new file mode 100644
index 000000000..6dd6d9955
--- /dev/null
+++ b/gitnexus/test/unit/u8-redos-resource-exhaustion.test.ts
@@ -0,0 +1,196 @@
+/**
+ * Regression tests for U8 — closes:
+ *   #186 js/redos             rust-workspace-extractor.ts
+ *   #187 js/redos             cobol-preprocessor.ts
+ *   #184 js/resource-exhaustion cross-impact.ts
+ *
+ * These tests import the production symbols directly. A previous shape
+ * dynamic-imported names that did not exist (`extractRustWorkspace` vs.
+ * the real `extractRustWorkspaceLinks`) and `??`-fell-back to inline
+ * regex copies, so the tests stayed green even when the production
+ * fixes regressed. Static imports + named symbols make a regression in
+ * any of the three sites a hard test failure.
+ */
+import { describe, expect, it } from 'vitest';
+import { RE_SET_TO_TRUE, RE_SET_INDEX } from '../../src/core/ingestion/cobol/cobol-preprocessor.js';
+import { parseCargoPackageName } from '../../src/core/group/extractors/rust-workspace-extractor.js';
+import {
+  clampTimeout,
+  IMPACT_TIMEOUT_MIN_MS,
+  IMPACT_TIMEOUT_MAX_MS,
+} from '../../src/core/group/cross-impact.js';
+
+/**
+ * Time a single regex.exec call. Used by the linearity tests below to
+ * compute a 10k/5k ratio in addition to the absolute <500ms bound.
+ *
+ * Ratio assertions catch sub-exponential O(n²) regressions that fit
+ * inside the absolute cap on warm CI; the absolute cap catches
+ * catastrophic backtracking on cold CI. Two complementary signals.
+ */
+function timeRegex(re: RegExp, input: string): number {
+  // Reset regex.lastIndex for global/sticky regexes — ours are not, but
+  // be defensive in case future shape changes add the `g` flag.
+  re.lastIndex = 0;
+  const start = performance.now();
+  re.exec(input);
+  return performance.now() - start;
+}
+
+function timeFn<T>(fn: () => T): number {
+  const start = performance.now();
+  fn();
+  return performance.now() - start;
+}
+
+// Linear scaling is ~2.0× when input doubles; 3.0× allows generous
+// slack for CI-runner GC and tier-up jitter. An O(n²) regression on a
+// 2× input takes ~4× as long, well outside this bound.
+const LINEAR_RATIO_BOUND = 3.0;
+
+/**
+ * Minimum elapsed time (in ms) below which `performance.now()` ratios
+ * are dominated by scheduler jitter and become meaningless. When both
+ * timed runs come in below this floor, we skip the ratio assertion —
+ * the absolute <500ms bound still catches catastrophic backtracking,
+ * and the next CI run will measure higher absolute times that the
+ * ratio assertion can evaluate reliably.
+ *
+ * Calibrated empirically: a flake on macOS reported ratio 5.29×
+ * between two sub-millisecond measurements (~0.5ms vs ~2.6ms), both
+ * genuinely linear but indistinguishable from noise. 5ms is a
+ * comfortable floor where individual measurements are well-separated
+ * from the ~10-100µs `performance.now()` resolution band.
+ */
+const RATIO_MEASUREMENT_FLOOR_MS = 5;
+
+/**
+ * Assert linear scaling between two timed runs on inputs that differ
+ * by 2×. When measurements are too small to be reliable, the ratio
+ * assertion is skipped (the absolute bound still fires elsewhere).
+ */
+function assertSubLinearRatio(elapsedSmall: number, elapsedLarge: number, label: string): void {
+  if (elapsedSmall < RATIO_MEASUREMENT_FLOOR_MS && elapsedLarge < RATIO_MEASUREMENT_FLOOR_MS) {
+    // Both runs completed faster than the noise floor — the ratio is
+    // not meaningful. The absolute <500ms bound elsewhere in this
+    // describe block still pins linearity; we skip rather than risk a
+    // flake on a genuinely-linear implementation.
+    return;
+  }
+  const ratio = elapsedLarge / Math.max(elapsedSmall, 0.001);
+  if (ratio >= LINEAR_RATIO_BOUND) {
+    throw new Error(
+      `${label}: ratio ${ratio.toFixed(2)}× exceeds bound ${LINEAR_RATIO_BOUND}× ` +
+        `(small=${elapsedSmall.toFixed(2)}ms, large=${elapsedLarge.toFixed(2)}ms)`,
+    );
+  }
+}
+
+describe('cobol-preprocessor RE_SET_TO_TRUE — linear time on pathological input', () => {
+  it('matches in <500ms on 50k repetitions of "A OF A " AND 100k/50k ratio is sub-linear when measurable', () => {
+    // 50k/100k repetitions chosen so timings exceed the
+    // RATIO_MEASUREMENT_FLOOR_MS noise floor on typical CI hardware.
+    // Pre-fix nested-quantifier shape would be exponential here; the
+    // post-fix `.+?` shape is linear (~2× when input doubles).
+    const inputSmall = 'SET ' + 'A OF A '.repeat(50_000) + 'TO TRUE';
+    const inputLarge = 'SET ' + 'A OF A '.repeat(100_000) + 'TO TRUE';
+    const elapsedSmall = timeRegex(RE_SET_TO_TRUE, inputSmall);
+    const elapsedLarge = timeRegex(RE_SET_TO_TRUE, inputLarge);
+    expect(RE_SET_TO_TRUE.exec(inputSmall)).not.toBeNull();
+    expect(elapsedSmall).toBeLessThan(500);
+    expect(elapsedLarge).toBeLessThan(500);
+    assertSubLinearRatio(elapsedSmall, elapsedLarge, 'RE_SET_TO_TRUE');
+  });
+
+  it('still matches a normal SET ... TO TRUE statement', () => {
+    const m = RE_SET_TO_TRUE.exec('SET WS-FLAG TO TRUE');
+    expect(m).not.toBeNull();
+    expect(m?.[1]).toBe('WS-FLAG');
+  });
+});
+
+describe('cobol-preprocessor RE_SET_INDEX — linear time on pathological input', () => {
+  it('rejects in <500ms on 50k tokens with no valid suffix AND 100k/50k ratio is sub-linear when measurable', () => {
+    // Forces backtracking against the (TO|UP\s+BY|DOWN\s+BY) alternation
+    // — the richer pathological surface of the two regexes.
+    const inputSmall = 'SET ' + 'A '.repeat(50_000) + 'X';
+    const inputLarge = 'SET ' + 'A '.repeat(100_000) + 'X';
+    const elapsedSmall = timeRegex(RE_SET_INDEX, inputSmall);
+    const elapsedLarge = timeRegex(RE_SET_INDEX, inputLarge);
+    expect(RE_SET_INDEX.exec(inputSmall)).toBeNull();
+    expect(elapsedSmall).toBeLessThan(500);
+    expect(elapsedLarge).toBeLessThan(500);
+    assertSubLinearRatio(elapsedSmall, elapsedLarge, 'RE_SET_INDEX');
+  });
+
+  it('still matches a normal SET INDEX statement', () => {
+    const m = RE_SET_INDEX.exec('SET WS-IDX TO 5');
+    expect(m).not.toBeNull();
+    expect(m?.[1]).toBe('WS-IDX');
+    expect(m?.[2]).toBe('TO');
+    expect(m?.[3]).toBe('5');
+  });
+});
+
+describe('rust-workspace parseCargoPackageName — linear-time line walk', () => {
+  it('extracts the package name in <500ms on 100k blank lines AND 200k/100k ratio is sub-linear when measurable', () => {
+    // 100k/200k blank lines chosen so timings exceed the
+    // RATIO_MEASUREMENT_FLOOR_MS noise floor. Earlier 10k/20k pairing
+    // produced sub-millisecond measurements where scheduler jitter
+    // dominated and the ratio became meaningless (a real macOS run
+    // saw 5.29× between two genuinely-linear sub-ms measurements).
+    const cargoTomlSmall =
+      '[package]\n' + '\n'.repeat(100_000) + 'name = "myrepo"\nversion = "0.1.0"\n';
+    const cargoTomlLarge =
+      '[package]\n' + '\n'.repeat(200_000) + 'name = "myrepo"\nversion = "0.1.0"\n';
+    const elapsedSmall = timeFn(() => parseCargoPackageName(cargoTomlSmall));
+    const elapsedLarge = timeFn(() => parseCargoPackageName(cargoTomlLarge));
+    expect(parseCargoPackageName(cargoTomlSmall)).toBe('myrepo');
+    expect(elapsedSmall).toBeLessThan(500);
+    expect(elapsedLarge).toBeLessThan(500);
+    assertSubLinearRatio(elapsedSmall, elapsedLarge, 'parseCargoPackageName');
+  });
+
+  it('returns null when [package] section is absent', () => {
+    expect(parseCargoPackageName('[workspace]\nmembers = ["a"]\n')).toBeNull();
+  });
+
+  it('stops at the next section header (does not pick up a name= from a later section)', () => {
+    const toml = '[package]\nversion = "1.0"\n[other]\nname = "wrong"\n';
+    expect(parseCargoPackageName(toml)).toBeNull();
+  });
+
+  it('extracts the name from a normal [package] section', () => {
+    const toml = '[package]\nname = "real-crate"\nversion = "0.1.0"\n';
+    expect(parseCargoPackageName(toml)).toBe('real-crate');
+  });
+});
+
+describe('cross-impact clampTimeout — bounds user-supplied impact timeouts', () => {
+  it('rejects negative and zero timeouts, returning MIN', () => {
+    expect(clampTimeout(0)).toBe(IMPACT_TIMEOUT_MIN_MS);
+    expect(clampTimeout(-1)).toBe(IMPACT_TIMEOUT_MIN_MS);
+    expect(clampTimeout(-999_999)).toBe(IMPACT_TIMEOUT_MIN_MS);
+  });
+
+  it('rejects NaN/Infinity, returning MIN', () => {
+    expect(clampTimeout(NaN)).toBe(IMPACT_TIMEOUT_MIN_MS);
+    expect(clampTimeout(Infinity)).toBe(IMPACT_TIMEOUT_MIN_MS);
+    expect(clampTimeout(-Infinity)).toBe(IMPACT_TIMEOUT_MIN_MS);
+  });
+
+  it('caps very large timeouts at MAX (5 minutes)', () => {
+    expect(clampTimeout(999_999_999)).toBe(IMPACT_TIMEOUT_MAX_MS);
+    expect(clampTimeout(IMPACT_TIMEOUT_MAX_MS + 1)).toBe(IMPACT_TIMEOUT_MAX_MS);
+  });
+
+  it('passes through a reasonable timeout unchanged (truncated to integer)', () => {
+    expect(clampTimeout(30_000)).toBe(30_000);
+    expect(clampTimeout(30_500.7)).toBe(30_500);
+  });
+
+  it('floors below-MIN positive values to MIN', () => {
+    expect(clampTimeout(50)).toBe(IMPACT_TIMEOUT_MIN_MS);
+    expect(clampTimeout(0.1)).toBe(IMPACT_TIMEOUT_MIN_MS);
+  });
+});
diff --git a/gitnexus/vitest.config.ts b/gitnexus/vitest.config.ts
index 9330e86a5..862357668 100644
--- a/gitnexus/vitest.config.ts
+++ b/gitnexus/vitest.config.ts
@@ -60,6 +60,8 @@ export default defineConfig({
             'test/integration/augmentation.test.ts',
             'test/integration/staleness-and-stability.test.ts',
             'test/integration/lbug-lock-retry.test.ts',
+            'test/integration/lbug-open-retry.test.ts',
+            'test/integration/lbug-close-handle-release.test.ts',
             'test/integration/api-impact-e2e.test.ts',
             'test/integration/shape-check-regression.test.ts',
             'test/integration/java-class-impact.test.ts',
@@ -87,6 +89,8 @@ export default defineConfig({
             'test/integration/augmentation.test.ts',
             'test/integration/staleness-and-stability.test.ts',
             'test/integration/lbug-lock-retry.test.ts',
+            'test/integration/lbug-open-retry.test.ts',
+            'test/integration/lbug-close-handle-release.test.ts',
             'test/integration/api-impact-e2e.test.ts',
             'test/integration/shape-check-regression.test.ts',
             'test/integration/java-class-impact.test.ts',