mirror of
https://github.com/abhigyanpatwari/GitNexus.git
synced 2026-08-28 05:25:25 +00:00
7 commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
Shane Thurston Wijaya
|
fc885a4bf3
|
docs(claude-skills): bind repository and worktree identity in multi repo skills (#2981) | ||
|
Gergő Magyar
|
2be508e796
|
fix(mcp): stop scaling the detect_changes query with the diff's hunk count (#2915) (#2930)
* fix(mcp): map diff hunks to symbols without per-hunk OR conditions (#2915) `detect_changes` folded one `(n.startLine <= $hunkEndI AND n.endLine >= $hunkStartI)` pair per diff hunk into a single WHERE clause, one query per changed file. A machine-generated file (cache JSON, lockfile, golden fixture) diffs at thousands of hunks with `-U0`, and the expression tree that produces overflows LadybugDB's recursive evaluator copy on a TaskScheduler worker thread: a bare SIGBUS with no error output where secondary threads get 512 KB of stack (macOS), a swallowed 30s query timeout where they get more (Linux), which the CLI then printed as "No changes detected." with exit 0. Coalesce each file's hunks into sorted, disjoint ranges and run the overlap test in JS instead. Only ranges that overlap or abut are merged, so the union covers exactly the lines the raw hunks covered. Query text and parameters are now identical whether a file changed in 1 place or 100,000, and files are queried in batches of 100 rather than one full node scan each. Reproduced on Linux by running the engine with macOS-sized (512 KB) thread stacks: 2,500 hunks passed, 3,333 and 4,000 segfaulted — matching the reporter's macOS threshold table. After the change the same repo maps a 100,001-hunk diff in 2.1s with no crash. Also fixes a line-base mismatch the rewrite exposed: graph rows are 0-based (#2377) while git hunk lines are 1-based, so the raw comparison shifted every symbol one line up. An edit to a symbol's LAST line reported nothing changed — a one-line function whose body was edited was invisible to the pre-commit gate. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * fix(cli): say when a detect_changes result is partial (#2915) When a graph query fails, `detect_changes` swallows the error, sets `partial: true` and leaves the counts at zero (#2283). The CLI formatter never read that flag, so a degraded run printed "No changes detected." and exited 0 — the pre-commit safety gate reporting a clean bill of health for a check that did not complete. Print the partial note in both the empty and non-empty branches. Also restore the `Symbol` placeholder for rows whose label came back as an empty string: the changed-symbol mapping now keeps `''` instead of dropping it to undefined, so the formatter needs `||`, not `??`. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * refactor(mcp): bound the hunk→symbol query and simplify the overlap helpers (#2915) Cleanup pass over the #2915 fix. No change to which symbols detect_changes reports, except that a node matched by two changed paths is now reported once. * Push a per-file [lo, hi] span into the query. Coalesced ranges are sorted and disjoint, so a file's whole touched span is free, and the engine can drop the symbols outside it instead of shipping every row in the file across the native boundary. Measured on a 400-file batch against a 25k-node index: 546ms/13,870 rows before, 84ms/1,555 rows after, identical kept set. Depth stays constant (two comparisons per file, not per hunk), so #2915 cannot come back — the JS test still rejects symbols landing in the gaps between hunks. The struct-list parameter was verified against @ladybugdb/core 0.18.3 and 0.19.1. * Convert hunks into the graph's 0-based space once, at the point they are grouped, with the existing `toZeroBasedLine`. Every comparison downstream is then base-neutral, and `toDisplayLine` goes back to being what its doc says it is: an MCP response-boundary converter, not a filter input. * Deduplicate matched nodes by id. `ENDS WITH` is a plain string suffix, so a diff touching both `README.md` and `pkg/README.md` counted the same node twice (169 duplicates in 13,870 rows on a real 400-file diff). Pre-existing, free to fix now that the rows are shaped in one place. * Drop the positional `?? sym[N]` row fallbacks in this block. `executeParameterized` returns `getAll()` rows, which are alias-keyed objects, so the fallbacks were dead — and they coupled the mapping to RETURN column order, which is what made adding a column a renumbering exercise. * Build the path→hunks map in one pass, so "every value is coalesced" holds at every point rather than being repaired by a second loop. Simplify `coalesceHunks` (the length<2 branch and the sort tiebreaker changed nothing) and state `hunksOverlapRange` as a standard half-open lower bound. * Document `partial` in the detect_changes tool description. The CLI now prints it, but the MCP client — the main consumer of the pre-commit gate — was getting the flag as an undocumented raw key. * Tests: pin the query text as identical for a 1-hunk and a 3,000-hunk diff (replacing a magic length bound), pin the 0-based bounds parameter, pin the dedup, and fold two near-identical row mocks into one helper. Temp dirs now come from the shared pool helper, whose cleanup is per-directory and Windows-lock aware. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * feat(mcp): bound and batch the hunk→symbol query, and anchor its path match (#2915) Follow-up review pass on the #2915 fix, implementing every remaining finding. * Push a per-file `[lo, hi]` span into the query. Coalesced hunks are sorted and disjoint, so a file's touched span is free, and the engine drops the symbols outside it instead of shipping every row in the file across the native boundary. Measured on a 25k-node index, 400-file batch: 546ms/13,870 rows before, 84ms/1,555 after, identical kept set. Depth stays constant (two comparisons per file, not per hunk), so #2915 cannot return. The struct-list parameter was probed against @ladybugdb/core 0.18.3 and 0.19.1 first; the index-subscript form `$paths[i]` does not parse. * Anchor the path match: `n.filePath = b.path OR n.filePath ENDS WITH b.suffix` where suffix is the path with a leading separator. A bare `ENDS WITH` is a plain string suffix, so a diff touching `lib/a.py` also reported a symbol from an indexed `src/mylib/a.py` — a file the diff never touched. This is the form `explain` already uses. Pinned by an integration test against a real engine (it fails 3/3 with the un-anchored predicate). * Run batches a few at a time. `executeParameterized` checks a connection out of the 8-connection per-repo pool for the duration of a query, so parallel calls never share one — the same reason ~15 other queries in this file already run under `Promise.all`. `allSettled`, so one failed batch degrades the result to `partial` instead of discarding the batches that succeeded beside it. * Deduplicate matched nodes by id, and count `changed_files` as distinct paths: a path can appear twice in one diff (a rename reported alongside an edit). * Cap the listed symbols at 1,000 with `symbols_truncated: {listed, total}`. A repo-wide diff otherwise puts an unbounded array in one MCP payload — the CLI has `--limit`, an MCP client has nothing. Counts are never capped, so the risk level and the CLI's "... and N more" still see the true total. * Extract `chunk` / `mapBatches` / `LBUG_QUERY_BATCH_SIZE` into `core/lbug/query-batch.ts`. Every query built from a caller-sized array has this ceiling; the shape now has one name and the measured batch size is recorded where it is defined rather than in three constants under three names. * Move hunk grouping and the 0-based conversion into `coalesceHunksByPath`, at the parse boundary. `parseDiffHunks` stays faithful to git (1-based, like the `@@` headers it reads), consumers compare graph-native values, and the conversion is unit-testable instead of living in the backend. * Document `partial` and `symbols_truncated` in the detect_changes tool description — the MCP client is the main consumer of the pre-commit gate and was getting both as undocumented raw keys. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * refactor(core): batch every remaining repo-sized query list (#2915) `detect_changes` was not the only place building query text from a caller-sized array. `core/wiki/graph-queries.ts` interpolated the whole file list of a module into four `IN [...]` literals, growing the query with the repo — flat breadth rather than the nested depth that crashed #2915, but the same unbounded shape, and the one the repo's own `DELETE_FILES_CHUNK_SIZE` precedent already chunks elsewhere. All four now run one query per batch and merge in JS. The membership arms need care, and each is documented where it happens: * `getIntraModuleCallEdges` batches the caller arm only. A per-batch callee arm would drop a call from batch 0 to batch 2, both inside the module, so that predicate moves to JS against the whole set. Results are now sorted: the single-query form had no ORDER BY, and batch order would hand the entire 30-edge window `formatCallEdges` keeps to the first 100 files (#2787). * `getInterModuleCallEdges` keeps the SAME batch list in its `NOT` arm. That is sound — a file outside the module is outside every batch — and it preserves the null handling: `NOT null IN [...]` is null, so the original dropped edges to a node with no filePath, where a JS-only `!has(undefined)` would admit them. ORDER BY and LIMIT move to JS because a per-batch limit would cut rows before the cross-batch membership filter ran. * `getProcessesForFiles` keeps `LIMIT` inside the batch: `stepCount DESC, id` is a total order, so a process in the global top-N is in its own batch's top-N. Also adopt the shared `chunk()` at the hand-rolled slice loops in `lbug-adapter.ts`, `embeddings/http-client.ts` and `run-analyze.ts`. The loops whose index fed a progress callback or an error message use `chunk(...).entries()`, which removes the `i / SIZE` and `Math.floor(i / SIZE)` arithmetic rather than reproducing it. No batch size changed. One trap that survived tsc and is worth naming: after renaming a loop variable away from `chunk`, a leftover `chunk.length` silently resolved to the imported FUNCTION's arity, reporting `chunkSize: 1` for a 200-path batch. Only `lbug-query-importers-batch`'s exact-value assertion caught it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * refactor: name the line-base conversions and share the symbol line (#2915) The 0-based-graph vs 1-based-elsewhere rule was open-coded in five places with the reasoning living only in comments — the same rule that, applied by hand and skipped once, hid every last-line edit from `detect_changes`. * Add `toOneBasedLine` beside `toZeroBasedLine` in `ingestion/utils/line-base.ts` so the module owns both directions, and adopt it at the four CFG/PDG join sites in `pdg-impact.ts` and the two in `local-backend.ts`. This is NOT `line-display.ts`'s `toDisplayLine`, which is documented as a response boundary converter with an `undefined` passthrough; the joins need arithmetic, and the guards that produce `Number.NaN` for an absent line are kept verbatim. * `http-route-extractor.ts` probed graph spans with a bare `line - 1` and a 20-line comment. It calls `toZeroBasedLine` now; the `?? pick(line)` fallback arm is untouched, so which node is picked cannot change (the clamp differs only for a negative line, which no emitter can produce). * Extract `formatSymbolLine`: `detect-changes-format.ts` and `eval-server.ts` rendered the same `type name → filePath` line. One behavior note — the two were not byte-identical, and eval-server had no placeholder on `name`, so a definition with an empty name rendered the literal `undefined` and now renders `?`. Both `definitions[]` shapes set name from a graph row, so this is unreachable in practice, and printing `undefined` into LLM-facing output is the bug, not the intent. `||` (not `??`) in the placeholders is deliberate and documented: a node label can come back as an empty string and still needs the placeholder. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * perf(wiki): bind the module file list instead of splicing it into the query (#2915) The wiki's four `IN [...]` sites interpolated every file of a module into the query text, so the text grew with the repo — the shape that overflowed LadybugDB's recursive evaluator copy in `detect_changes`. The previous commit chunked them, which worked but cost real complexity: the callee arm had to leave Cypher and be re-implemented in JS, DISTINCT had to be re-established across batches, and ORDER BY/LIMIT had to move to JS so a per-batch window could not cut rows the cross-batch filter still needed. Binding the list as a parameter removes the reason for all of it. The text is constant at any list length, and measured against a real index a bound list is ~3x faster than the equivalent literal (5,000 items: 139ms vs 459ms; 20,000: 598ms vs 1,686ms). Every predicate goes back into Cypher, including the `NOT ... IN` arms whose null handling is load-bearing — `NOT null IN [...]` is null, so a callee with no filePath is dropped by the engine, where a JS membership test would have admitted it. Verified on this repo's own index: a 2,000-path bound list returns 14,856 rows in 877ms. Also collapses the per-process step query into one grouped `p.id IN $ids` fetch — 105ms to 13ms for 20 processes — and drops `fileListLiteral`, `callEdgeKey`, `compareProcessHeaders` and the batching loops with it. `compareStrings` was a byte-identical re-roll of `compareCodeUnits` (src/lib/utils.ts), including its #2787 rationale; it now calls the shared one. Intra-module edges are sorted where the original had no ORDER BY: `formatCallEdges` keeps only the first 30, and an unordered cut keeps a different subset per machine. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * refactor(core): one home for batching, and a backstop for the shape that crashed (#2915) `chunk` moves to `src/lib/utils.ts`, the repo's generic-utility home: it is an array helper, and leaving it in `core/lbug/query-batch.ts` made an HTTP embedding client import batching from the graph-DB namespace. `query-batch.ts` keeps what is actually about queries — the measured `LBUG_QUERY_BATCH_SIZE`, the concurrency helper, and the ceiling — and now documents the preference the wiki change proved: bind the list as a parameter first, chunk only when you cannot. `mapBatches` becomes `mapConcurrent`: nothing about it is batch-specific, and it now has non-query callers. Its body is a per-item try/catch plus `Promise.all`, so ordering comes from the primitive rather than from unwrapping a settled union. The wave barrier stays — measured against a rolling window it is 538ms vs 532ms on a 1,000-file diff, whose per-batch times spread only 1.35x. Adopted at the loops that were still hand-rolled: `file-hash.ts`, `cluster-enricher.ts` (its progress callback now accumulates `batch.length` instead of clamping an index), `filesystem-walker.ts` and `language-config.ts` (wave scheduling with `allSettled`, which is exactly `mapConcurrent`). Deliberately not adopted, each for a stated reason: the analyzer-identity probe runs as a standalone `node -e` script with no module resolution; the embedding sub-batch loop slices two parallel arrays and breaks early; `walkRepositoryPaths` reports progress from inside each wave, which `mapConcurrent` cannot express. `warnIfQueryTextUnbounded` is the backstop: #2915 died in native code with no message, and a query built by concatenating a caller-sized list is the shape that gets there. Wired at both execution chokepoints (`pool-adapter`'s `executeParameterized`, `lbug-adapter`'s `executePrepared`/`streamQuery`; their `executeQuery` siblings delegate and are covered once). It never throws — a long query the engine can actually run must not start failing on a heuristic — and it is deliberately absent from the raw write path, where a node's `content` is inlined and a large source file would warn legitimately. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * refactor(mcp): name the path-match rule, and key detect_changes by node id (#2915) * `path-predicate.ts` names the three ways a caller's path can match a stored `filePath` — `exact`, `pathSuffix`, `fragment` — instead of each call site copying whichever idiom its neighbour used. A bare `ENDS WITH` is a plain string suffix, which is how a diff touching `lib/a.ts` came to report a symbol from `src/mylib/a.ts`; the loose `CONTAINS` sites are loose ON PURPOSE (a user hint of `src/mcp` should match a directory fragment), and naming the modes is what lets a call site choose rather than inherit. * `detectChanges` kept four structures over one row set — an array, a dedup Set, an id list and an id→name Map — that had to stay in sync by hand. One id-keyed Map is all of them; insertion order is preserved, so every output is byte-identical. * `symbols_truncated: {listed, total}` becomes `truncated: true`, the key `explain`/`pdg_query`/`trace` already use. The true total was always in `summary.changed_count`, so the nested object said nothing the existing vocabulary could not. * `GraphLineRange` is now a distinct type from `DiffHunk`: they carry the same two fields in different bases, and mixing them IS #2377. The name means a 1-based hunk cannot reach `hunksOverlapRange` without a conversion between. * `coalesceHunksByPath` accumulates raw ranges and coalesces once per path rather than re-sorting on every occurrence. * `chunk` adopted at this file's own five loops — the point of extracting it — including two locals named `chunk` that shadowed the import. That shadowing is not cosmetic: it is how a leftover `chunk.length` silently became the function's arity earlier in this branch. One bug caught by the real-engine integration test and worth naming: Cypher comments are `//`, not `--`. A `--` comment inside the query string made LadybugDB reject the whole query at PREPARE, which `detect_changes` swallows into `partial` and renders as "No changes detected." Every mocked unit test passed. Prose stays out of query strings. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * refactor(test): share the git-repo bootstrap, and move the shared formatter out (#2915) `formatSymbolLine` lived in `detect-changes-format.ts` but is rendered by `eval-server`'s query formatter too, so a `query` formatter imported from a `detect_changes` module. It moves to `src/cli/format-symbol.ts`; both callers import it from there. The `||`-not-`??` fallbacks stay documented — a node label can come back as an empty string and still needs its placeholder. `test/helpers/temp-git-repo.ts` gives `initGitRepo(dir, identity?)` and `commitAll(dir, message)` to the ~10 test files that hand-rolled the same `git init -q` + two `git config` + `add -A` + `commit` sequence. It takes a directory and never owns one, matching `temp-dir-pool.ts`'s split of lifecycle from seeding; the identity is a parameter because the existing consumers genuinely disagree about it, and each keeps exactly what it configured. Four files stay hand-rolled for stated reasons — pinned author dates for a deterministic digest, remote handling, `--allow-empty`, and the `-c key=value` form that never persists to the repo. Test trims: the `formatSymbolLine` fallback cases collapse into one `it.each` table (the case pinning that BOTH consumers emit the helper's exact line stays — no table row can express it); two `line-base` cases that were compositions of their neighbours go; and `detect-changes-path-anchoring` runs its `detect_changes` call once in `beforeAll` instead of three times, keeping the three named failures. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019d85r7TrMYjWbTUT3pXccS * perf(mcp): filter the batched hunk query before the engine materialises (#2915) `UNWIND $bounds AS b MATCH (n) WHERE …b…` compiles to a CROSS_PRODUCT whose build side is a RESULT_COLLECTOR over the whole filtered node table: only the `n`-only predicates get pushed below the accumulate, so neither the anchored path match nor the [lo, hi] span could reduce the scan. Measured at 1M nodes: +242 MB for one batch and +922 MB for the four concurrent ones, paid even for a one-file diff — and at a 268 MB buffer pool the query died with `Buffer manager exception` where the old per-file query completed, landing in `partial:true` + `changed_count:0`, the #2915 false clean by another route. Adding the batch-wide, `b`-free disjunction as a redundant leading conjunct lets the planner push it below the accumulate: EXPLAIN now shows it as FILTER[2] directly under SCAN_NODE_TABLE[0]. It is a provable superset of the correlated predicate, so it cannot drop a row the correlated filter keeps. 10x less memory, ~20% faster, identical result sets. Also in detect_changes: - Sort rows on (filePath, startLine, id) before the 1000-symbol cut. The cut was slicing engine row order — measured 5 distinct orders across 8 runs on one connection, the #2787 class this branch fixes 200 lines away in the wiki. - Chunk `symIds`, the one caller-sized list left unbatched: 500k ids measured 4.0 GB RSS. Binding keeps the query TEXT constant, which is all the unbounded guard measures, while the bound VALUE stayed repo-sized. - Prefer exact path equality and widen to the anchored suffix only for paths that matched nothing, so a root README.md stops reporting pkg/*/README.md. - Report `risk_level:'unknown'` rather than 'low' when a query was swallowed. A degraded pre-commit gate must not read as an all-clear. - Pass --no-ext-diff --src-prefix=a/ --dst-prefix=b/. `diff.noprefix` in a user's gitconfig makes git emit `+++ f.py`, which parseDiffHunks cannot match, so every run printed "No changes detected." and exited 0 before any query ran. A diff that parses to zero files now raises `partial` instead of the clean branch. - `labels(n)`, not `labels(n)[0]`: labels() returns a scalar string here, so the subscript was always '' and `type` never carried a label. - Validate IMPACT_MAX_CHUNKS. The chunk() adoption turned an entry condition into an exit condition, so a non-numeric value ran every chunk instead of none. - Record why four-way concurrency is safe here, and scope the arm64 sequential comment to the query it was written for (#496). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * fix(cli): fail the detect_changes gate instead of exiting 0 when it degrades (#2915) The secondary half of #2915 was that a swallowed query failure printed "No changes detected." and exited 0, so a shell pre-commit gate passed on a broken analysis. This branch added the PARTIAL text. It did not change the exit status, so `gitnexus detect-changes && git commit` still proceeded. `detectChangesCommand` passed a STRING to `output()`, and `output()` sets a failing code only for an OBJECT carrying `error` — under a comment calling itself "the one place that keeps scripted callers honest". A string never matches, so this command opted itself out of the only mechanism the file provides. It was broader than `partial`: the formatter also renders a backend `{error}` payload as text, so hard failures exited 0 too. Fixed narrowly in `detectChangesCommand`, following the object-first shape `checkCommand` already uses, rather than widening `output()`'s shared contract — every one of its other seven callers already passes an object and is unaffected. One code for both `error` and `partial`: `&&` only distinguishes zero from non-zero, and a softer code for `partial` would invite `|| [ $? -eq 2 ]` exemptions that reopen exactly this hole. `truncated` deliberately stays exit 0 — only the listing is capped, while the counts and risk are computed over the full set, so the verdict is sound and failing on it would fire on every large-but-healthy diff. Also wires `truncated` through the formatter, which this branch had left as a producer-only flag while `partial` went end to end, with the note in both locales and no count of its own so the existing "... and N more" line stays the sole numeric report. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * fix(wiki): restore step order and symbol labels, and cut the edge list in Cypher (#2915) Found by running the queries against a real engine, which nothing did before: this branch's regrouped `withSteps` returned step traces OUT OF ORDER. `ORDER BY pid, r.step` combined with `WHERE p.id IN $ids` silently drops the second sort key — `proc_1_incrementalupdate` came back 2,7,1,3,4,5,6. `ORDER BY step` alone is correct, and so was the pre-branch per-process query, so this was introduced by the batching. `formatProcesses` prints "${s.step}. ${s.name}", so every module and overview page was getting scrambled execution traces. The mocked suite passed 112/112 before and after. `labels(x)[0]` is always the empty string: labels() returns a scalar string and the subscript is 1-based over its characters ([1] is "F"). `prompts.ts` renders "${s.name} (${s.type})", so all 5,027 exported symbols reached the LLM as "name ()". `getIntraModuleCallEdges` shipped every edge to use 30 — measured 18,299 rows and 851 ms with all 2,079 paths bound, against 30 rows and 94 ms with ORDER BY + LIMIT in Cypher, which the sibling `getInterModuleCallEdges` twenty lines below already did. The determinism fix (#2787) was right; the placement was not. `compareCallEdges` goes with it — it was intransitive when a name was null or empty, so `Array.sort` was input-permutation dependent, i.e. the nondeterminism it was added to remove. Deletes the positional row ABI this branch newly documented. The vendor declaration is `getAll(): Promise<Record<string, LbugValue>[]>` — string keys only — and `row[0]` probes back `undefined`; the same PR deleted ~30 identical fallbacks from local-backend.ts. They were already stale here: `withSteps` prepends `p.id AS pid`, so `toProcessStep` was reading the pre-branch layout. Rows are now typed by alias, so renaming an `AS` is a compile error. `??` for `||` so a step of 0 or an empty label keeps its own value. Tests: a real-engine integration suite covering all seven exported queries (PREPARE included — the trap that shipped a `--` comment on this branch), and the four holes that let the ordering bug through — a vacuous order assertion, a LIMIT never reached by a 2-edge fixture, a fake that returned rows pre-ordered and ignored ORDER BY, and a hardcoded `type: 'Function'` that hid labels(). The step-ordering fixture is empirically sized: 2 processes never reproduced the bug, ~400 step edges was intermittent, 710 (20 processes x 26-45 steps) hit 11 of 11 runs. Seeded descending and interleaved so no grouping looks sorted by accident. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * refactor: put the shared helpers where their callers are, and make their contracts true (#2915) `mapConcurrent` moves to lib/utils.ts beside chunk(). Nothing about it is query-specific and it already had filesystem callers, while its docstring justified concurrency safety through the per-repo connection pool — an argument that does not apply to fs.readFile. This is the precondition the branch's own commit message stated ("it now has non-query callers") and then did not apply. LBUG_QUERY_BATCH_SIZE and warnIfQueryTextUnbounded genuinely are query-specific and stay. `pathMatch`/`PathMatchMode` deleted: zero callers, and none of the three sites its docstring cited were migrated, so the tree carried the abstraction and the copies it was written to replace. `pathSuffixOf` stays and the module now documents the anchoring rule it actually implements. Contracts that were not true: - QUERY_TEXT_CEILING_BYTES was compared against `cypher.length` — UTF-16 code units, not bytes — so non-ASCII query text was undercounted and the reported KB was wrong. Buffer.byteLength now, behind a `length * 3 <= ceiling` early return so only text over ~21 KB pays for the count. - chunk(items, NaN) returned [[]], against a docstring promising never to return an empty slice, and mapConcurrent's Math.max(1, NaN) propagated it — which would have resolved [] for non-empty input with no error, read as "no results" by every call site. - GraphLineRange claimed a 1-based hunk could not reach hunksOverlapRange without a conversion, but it was structurally identical to DiffHunk so tsc accepted one with no diagnostic, and coalesceHunks<T extends GraphLineRange> actively laundered the base while its accumulator was still DiffHunk[]. The useless generic is gone and a one-line phantom on each interface makes the claim real; a bare {startLine, endLine} literal still satisfies both, so no construction site needs a cast. Pure deletions no longer vanish. A -U0 deletion emits `+N,0`, which parseDiffHunks dropped, so the file survived with no hunks, no query ran, and detect_changes reported `changed_files:1, changed_count:0, risk_level:'low'` — "No changes detected." for a commit that deleted a function. A unified diff spells an empty range as the line before it, so the anchor is line N alone: a symbol containing the deleted text also contains N, while extending to N+1 would claim a symbol that merely starts after the gap — the widening coalesceHunks guarantees it never does. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * docs: say that a partial or truncated detect_changes is not a clean gate (#2915) The gate itself now fails loudly, but the instructions every agent reads still described a zero as a result. Fixed at the source: AGENTS.md's gitnexus block is generated from a template in cli/ai-context.ts and injected into every user's repo, so the sentence goes there and AGENTS.md/CLAUDE.md are regenerated through the real code path (which also picks up a pre-existing `analyze --index-only` drift the committed docs were behind). That block is under a test-enforced size cap with 30 characters of headroom, so the 144-character clause was paid for in the same currency: the header exhortation, which the Always Do list restates as MUSTs with commands, and a verbatim repeat of the detect-changes command in the regression-compare example. 3549 of 3552. Worth noting for whoever adds the next line — #2899 replaced an absolute cap with a 0.65 ratio to let "a legitimate clause fit without ceremony", but set the ratio flush against the block's then-current size, so it is a ratchet with no ratchet. The canonical block does not make the skills redundant: three of the four install channels ship skills without touching AGENTS.md, --skip-agents-md does the same in-repo, and a user-trimmed gitnexus:keep block legitimately has no Always Do section — in those repos the skill file is the only carrier. Precedent agrees: the risk:UNKNOWN rule is deliberately carried in both places. So one sentence each in gitnexus-work (the commit gate), gitnexus-impact-analysis (beside the UNKNOWN paragraph) and gitnexus-refactoring, whose post-hoc "verify only expected files changed" is the worst of the three because a degraded result makes it vacuously pass. gitnexus-taint-analysis is left alone: its audience is always inside this repo, where the canonical block loads. All copies mirrored to npm, plugin and cursor. The cursor copies are condensed checklists rather than byte-mirrors, so they carry the equivalent note placed where it governs every detect_changes line in the file — and nothing tests that, since standard skills are fragment-checked rather than byte-compared. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * refactor: break the seven small import cycles gitnexus check reports (#2915) `check` reported 11 cycles. Five are paths inside a single 257-file strongly connected component in core/ingestion (call-extractors / cfg visitors / utils/ast-helpers), with a second 26-file component behind it — fixing those paths would only make check print different ones, so both are left for their own PR. This closes the seven that are genuinely separable, taking the graph from 9 strongly connected components to 2. Six of the seven were one value import plus one `import type` edge. tsconfig sets neither verbatimModuleSyntax nor isolatedModules, so those edges erase entirely — the cut is a graph and readability change with no emitted-JS difference. Each moved type went to a leaf module, with a re-export left behind only where an importer outside the change actually needed it: - cli/ai-context <-> cli/skill-gen: GeneratedSkillInfo -> cli/generated-skill.ts. One importer, no package export surface, so a clean move with no re-export. - cli/analyze-config <-> cli/analyze (+core/run-analyze): AnalyzeOptions -> cli/analyze-options.ts. Re-export kept because a test imports it from analyze.js. run-analyze needed no edit — cutting the one type edge collapses the 3-file component into a DAG. Its own same-named AnalyzeOptions is a different interface and was deliberately not merged. - ingestion/import-resolvers/types <-> ingestion/language-config: type-only in BOTH directions, so it had no runtime existence at all. ImportConfigs has no importers outside the pair and is the return type of loadImportConfigs, so it moved into language-config. Side effect worth having: the shared resolver types module no longer names a single language, which is an AGENTS.md rule for core/ingestion shared pipeline code. - ingestion/di-extractors barrel <-> spring: DiResolver and the two match types -> di-extractors/types.ts, following the import-resolvers/types.ts precedent. - scope-resolution/walkers <-> workspace-index: WorkspaceResolutionIndex -> workspace-index-types.ts. Re-export is load-bearing — 9 src importers, 4 test files, and a dynamic import() at contract/scope-resolver.ts. Moving the value isClassLike instead was rejected: ~15 value importers, and it is documented as a pair with isShapeLike. - server/analyze-worker <-> analyze-worker-core: the WorkerMessage protocol -> analyze-worker-protocol.ts, a declarations-only leaf. storage/branch-index <-> storage/repo-manager was the one genuine two-way runtime cycle: branch-index called getStoragePaths/loadMeta, repo-manager used branchSlug/BRANCHES_DIR. branch-index's header conceded the cycle and argued it was ESM-safe because neither side calls across at module-evaluation time — a guarantee resting on call ordering rather than structure. Folding resolveBranchPlacement back the other way does not help, because BranchSummary.stats is typed RepoMeta['stats'], so RepoMeta had to move either way. Extracted storage/repo-meta.ts, a leaf importing only fs and path, holding the metadata read primitives; repo-manager re-exports the public names so all 54 RepoMeta and 50 loadMeta importers are untouched. The moved block diffs byte-identical against HEAD. Verified beyond typecheck, because the worker entrypoint is the risky part and nothing in the suite forks it: emitted analyze-worker.js still contains exactly one runtime import, and forking the real worker over IPC boots it through entry -> core -> protocol -> terminal-claim. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * refactor: apply the reuse, simplification, efficiency and altitude cleanups (#2915) The one that mattered: the degradation exit code was fixed at the wrong depth. `output()` has never inspected `partial` — it tests `error` only — so putting the check in `detectChangesCommand` left every other tool exiting 0 on a degraded run. `partial` is cross-tool vocabulary: query (enrichmentDegraded || ftsPartial), impact (!traversalComplete, perSymbolEnrichmentCapped) and the mode:'pdg' envelope all emit it. A truncated impact traversal returns a short caller set and an under-ranked risk, then exits 0 — so `gitnexus impact … && <edit>` proceeds, in the tool AGENTS.md makes a MUST gate before every edit. The justification also cited checkCommand as precedent, but checkCommand passes STRINGS too — it was the second command already hand-rolling around this gap, while output()'s docstring called itself "the one place that keeps scripted callers honest". output() now takes an optional renderer and fails on error OR partial; two hand-rolled sites go away and three tools are covered instead of one. truncated stays exit 0 (only the listing is capped) and checkCommand's cycleCount policy stays put. Efficiency, all re-measured on the 25k-node index: - The process lookup was chunked with LBUG_QUERY_BATCH_SIZE, calibrated for the opposite query shape — that constant is for a whole-node-table scan where more items amortise the scan, while this is an `id IN $ids` probe where round trips dominate. 20k ids: 617ms at 100, 261ms at 1000. New LBUG_ID_PROBE_BATCH_SIZE, documented against its sibling so they cannot be re-merged. This also settles the older "chunking this query is a regression" measurement — that was chunk=100. - The sort comparator re-coerced fields ChangedSymbolRow already types, O(n log n) redundant conversions (+31-38%). Row shape probed directly: alias-keyed, no positional keys, numeric columns are JS numbers. - exactlyMatchedPaths built two throwaway arrays; one loop instead (40k rows 11.4ms -> 4.5ms). - The integration fixture seeded 710 step edges one round trip at a time; one UNWIND instead. File wall time 6.91s -> 3.63s. Fixture size unchanged — its docstring records the threshold below which the bug stops reproducing, and the mutation check still fails 3/3 when ORDER BY step is reverted. Reuse and simplification: - CALL_EDGE_LIMIT existed in four places; its own docstring predicted the drift it then caused. prompts.ts owns it now — it is a zero-import leaf so the direction cannot cycle, and had graph-queries.ts owned it the four suites that vi.mock that module would have left slice(0, undefined), silently returning every edge in exactly the tests meant to police the cap. - Six dead positional row fallbacks survived the rewrite in the loop this branch re-indented, in the same PR that deleted the identical ABI from graph-queries.ts. - Two test files independently modelled the same labels() scalar-string quirk. Deleted the wiki one — the file's own new header says semantics belong in the real-engine test — and kept projectTypeColumn, the only instrument that can see the bug for the detect_changes query. - makeRepo onto the shared git bootstrap (the eleventh copy of the sequence the helper was extracted to own), the duplicate diff-args unwrapper merged into test/helpers, hand-rolled comparators onto compareCodeUnits, real-timer sleeps replaced by wave-released promises with a strengthened per-wave assertion. - Re-exports trimmed to what is actually imported, a cross-reference this branch invalidated by moving mapConcurrent, and a "~20% faster" claim that does not survive at real index sizes (1-9%; the 10x memory win does). Also adds the drift guard the new doc text lacked: fragment coverage for the partial/truncated paragraph in every skill copy and in the managed AGENTS.md / CLAUDE.md block. Falsifiability checked — none of those fragments exist at the merge base. Not done here, deliberately: 27 live labels(x)[0] projections remain across impact/context/query/trace and MCP resources, with four load-bearing workarounds that have begun depending on each other and one that fabricates rather than degrades. That is a semantic change to five agent-facing tools and wants its own PR, scoped to delete the workarounds too. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * fix(cli): restore the detect-changes subcommand in the regression example (#2915) Caught by the gitnexus-check bot on the PR. The regression-review fallback in the injected mandate rendered as `--scope compare --base-ref "main" --repo .` with no command, so anyone copying it invokes the runner with an option as its first argument. Self-inflicted, and by exactly the mechanism flagged when it landed: the block is under a test-enforced size cap (#856) that had 30 characters of headroom, so adding the partial/truncated clause required paying for it, and the 38-character "repeat" that was dropped turned out to be the subcommand rather than a repeat. Paid for the restoration out of the clause instead — both parentheticals are gone, since `partial` and `truncated` are already defined in the tool description this text points at. Block is back under the cap at 3548/3552. Notably the cap has now been raised four times (2700 -> 2900 -> 2950, then 0.55 -> 0.65) each with the argument that the new line is load-bearing, and it has now also caused a user-facing defect. It is not functioning as a budget. Left at 0.65 here rather than making it five: moving the threshold to fit one's own text is how it got here. Worth restructuring separately. The fragment guard added a commit ago caught the rewording immediately, which is what it is for; its fragments now pin the two policy claims rather than the prose around them, since that prose is what gets re-trimmed under the cap. Also verified and NOT changed: the bot's other error, that detect_changes compares 1-based hunks against 0-based graph lines. `bounds` is built from `coalesceHunksByPath`, which applies `toZeroBasedLine` to both ends at the grouping boundary, and both a mocked and a real-engine test pin an edit landing on a symbol's last line. The bot read `parseDiffHunks` in isolation. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * fix(core): reject a fractional chunk size, and stop the truncation note overclaiming (#2915) All five from the gitnexus-check bot's pass on the previous push; two were introduced by the cleanup round that preceded it. `chunk` guarded with `Number.isFinite`, which admits a fractional size — and that one does not fail, it DUPLICATES. `slice` truncates its indices while `i` does not, so size 1.5 yields slice(0, 1.5) = items 0-1 then slice(1.5, 3) = items 1-2, putting item 1 in two batches; a caller batching a query would send it twice. A size is a count, so `Number.isInteger`. Unreachable today (every caller passes a constant) but the guard existed precisely for the unreachable case, and the NaN half of it was already there. `mapConcurrent`'s per-item degradation contract had a hole: `onError` is caller-supplied and was invoked outside a try, so a throwing reporter rejected `settle`, rejected the whole `Promise.all` wave, and discarded the neighbouring successes the function exists to preserve. Reporting a failure must not become one. The CLI truncation note asserted "the counts and risk level still cover all of them", which is true only when `truncated` fires alone — with `partial` the counts are summed from the batches that succeeded. It now varies: a distinct string when both flags are set, saying the counts are a lower bound. This is the same claim already corrected in the tool description; the CLI text still had the old one. The di-extractors contract docstring claimed the barrel re-exports everything from it. That stopped being true when the re-export was trimmed to what is actually imported, one commit earlier. The real-engine wiki test claimed to prepare "every exported query" and omitted `getInterModuleEdgesForOverview`, which `generateOverview` calls. Added — it aggregates in JS over `getInterFileCallEdges` rather than issuing its own Cypher, so the note says why it is in a prepare test. Verified and NOT changed: the bot's other error, that detect_changes compares 1-based hunks against 0-based graph lines. `bounds` is built from `coalesceHunksByPath`, which converts both ends at the grouping boundary (storage/git.ts), and two tests pin an edit landing on a symbol's last line. The remaining seven findings are changed-symbol heads-ups with no signature change; their callers' suites are green. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv * fix(mcp): make the IMPACT_MAX_CHUNKS fallback actually fire (#2915) The validation added earlier this branch used `Number.parseInt`, which takes the numeric PREFIX: '1.5' parses to 1, satisfies `Number.isInteger`, and silently caps enrichment after a single 100-item batch — the opposite of the fallback the comment beside it promised. `Number` instead, so a fractional value is rejected and falls back to 10. The emptiness check is load-bearing rather than defensive: `Number('')` is 0 and 0 is a legitimate value here (enrich nothing), so an UNSET variable would otherwise mean "enrich nothing" rather than "use the default". Behaviour table, old vs new: '1.5' 1 -> 10 (the bug), and undefined/''/' '/ '10junk'/'-2'/'all' -> 10, '0' -> 0, '3' -> 3, ' 5 ' -> 5 all unchanged. So the only case that moves is the reported one. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0139ycbqnAorkJGQaZQXUNuv --------- Co-authored-by: Gergo Magyar <gergomagyar0@gmail.com> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> |
||
|
DuduPhudu
|
fa31a7d824
|
fix: close the nine follow-up review findings from #2856 (routes, receiver typing, truncation honesty) (#2899)
* fix(typescript): a type parameter shadows a declared type of the same name (W2-8)
First item of wave 2, promised to the reviewer on #2856.
`export function unwrap<Result>(value: Result): Result` names the PARAMETER, not
the `interface Result` beside it — tsc resolves both annotations to the
parameter. The type-reference capture that makes a contract answerable ("what
breaks if I remove this field?") had no notion of a parameter binding, so every
annotation mentioning `Result` inside `unwrap` minted a `USES` edge into the
interface, at the same confidence as a real consumer and indistinguishable from
one. Measured on the new fixture: `unwrap` produced TWO false edges while the
genuine consumer produced one.
Blast radius is every generic whose parameter name collides with a declared
type, and the colliding names are ordinary choices for both: `Result`, `Key`,
`Value`, `Item`, `Node`, `Options`, `Config`, `Props`, `State`, `Response`.
TWO HALVES, and the first is why upstream's fix could not reach this. #2833
introduced `bindsTypeParameter` for the CALL-receiver path, where a workspace
`class T` was answering for `<T>`. Reusing it here changed nothing at first, and
the reason is its own documented contract: `@declaration.type-parameters` was
captured for class/interface declarations ONLY, so a generic FUNCTION recorded
no parameter list and the predicate correctly returned false — absence is not
evidence. The data was missing, not the logic. So:
- TYPESCRIPT_SCOPE_QUERY now captures type parameters on `function_declaration`,
`generator_function_declaration` and `type_alias_declaration`;
- the graph bridge consults `bindsTypeParameter` before emitting `USES`.
Both are load-bearing — removing either one fails the fixture.
The fixture carries two controls, because the obvious wrong fix is to stop
emitting: a genuine consumer of the interface must still link, and a generic
whose parameter does NOT collide must still link its real reference. Both are
asserted, and the "genuine consumer" case is asserted FIRST so the absences
below it cannot pass vacuously.
SCHEMA_BUMP 53 -> 54: parse-time capture change. A warm cache replays defs with
no parameter list, so the guard reads nothing and the feature is inert while
looking implemented.
Capture fingerprint re-baselined with justification. NO NEW CAPTURE NAME —
diffing the capture-name sets against the wave-1 branch returns empty; the tag
existed and now fires on more declarations. capture_groups_fp 2338 -> 2371,
fixture_count 151 -> 152, scaling 1.06 < 1.5, and JavaScript's fingerprint does
not move at all, which is the check that this is the TS declaration rules rather
than something broader.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(analyze): close the four false-success paths in the graph-write-collapse guard (W2-6)
Second wave-2 item, promised on #2856. All four were reported; all four
reproduced by reading the code they name.
(a) A SAME-COMMIT RE-RUN REPORTED SUCCESS FOREVER. Every other meta-driven
trigger — schema fingerprint, PDG mode, runner identity, CJK segmentation,
embedding dims — has a block that forces a rebuild before the
`alreadyUpToDate` fast path. `graphWriteCollapsed` had none; `grep -rn` found
writes and no reads. So the one state meaning "most of your edges are gone"
was the one state that repaired itself only if the user happened to pass
`--force`. Now forces a full rebuild, and forcing is right rather than merely
re-running: the persisted graph disagrees with what the pipeline produced, so
an incremental pass over unchanged files would write nothing and re-stamp the
same broken index as fresh.
(b) AN INCREMENTAL RE-RUN ERASED THE STAMP. `saveMeta` is a full atomic
overwrite, and the field was spread in only when the CURRENT run had a
verdict. `undefined` meant two different things at that site — "full run, no
collapse" (a positive all-clear) and "incremental write, not comparable" (no
opinion) — so the second case silently dropped `graph-write-collapsed` from
meta.json while the edges were still missing. Now three-way: stamp on
detection, CLEAR on a healthy full run, CARRY FORWARD when there is no
verdict. That is the shape `branch: branchLabel ?? existingMeta?.branch` two
lines away had all along.
(c) THE SERVER PATH NEVER CONSUMED IT. `analyze-worker-ipc.ts` projects the field
"so a server-side caller sees the same degraded outcome the CLI does" — but
nothing read it, so the comment described an intention and every collapsed
run reported `complete` to the UI and to every API consumer. Now reports
`failed` with the counts and the remedy, matching the CLI, which prints
`Repository indexed INCOMPLETELY` and exits non-zero. A consumer that reads
"complete" will query the index and get confident wrong answers.
(d) --pdg ROWS MASKED TOTAL STRUCTURAL LOSS. `expected` counts the in-memory
graph plus the streamed STRUCTURAL manifest; the streamed PDG layers never
enter `graph.relationshipCount`. But `persisted` was `stats.edges`, a count
of EVERY `CodeRelation` row, and PDG writes into that same table. With 1,000
structural edges expected and 4,000 PDG rows persisted, losing every
structural edge still read `persisted = 4000`, cleared the ratio, and stayed
silent — on exactly the large repos `--pdg` is used for.
Worth recording that the OBVIOUS fix does not work. Padding `expected` with
the PDG rows makes the two universes match but leaves the ratio judging a
minority population: 4,000 of 5,000 still clears 0.5. I wrote that first, and
the test I wrote to prove it failed. Only comparing structural against
structural asks the question the check exists to ask, so `getLbugStats` gains
a `structuralEdges` count excluding `PDG_EDGE_TYPES`. `TAINT_PATH` is
deliberately NOT in that set — it is a whole-program Function→Function edge
persisted by the normal emit, so it is structural and stays counted on both
sides.
`index-freshness-graph-collapse.test.ts` had pinned the masking as correct
(`detectGraphWriteCollapse(1000, 4000)` → undefined, "PDG layers write into
the same table, so persisted > expected is normal"). True about the table,
and it licensed the hole. Replaced with the case that matters and a note on
why the fix is at the caller.
The new `structuralEdges` assertion in `lbug-core-adapter` is there because the
failure mode is silent: the query sits in a try/catch that yields `undefined`,
and `undefined` makes the collapse check decline to compare — so a typo in the
Cypher would throw nothing, fail nothing, and switch the guard off. Verified
against a real LadybugDB and mutation-checked by breaking the query.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(processes): make process selection insertion-order invariant (W2-5)
Third wave-2 item. Reproduced before fixing: two equal three-step flows with
`maxProcesses: 1` select `handleAlpha`; inserting the identical nodes and CALLS
edges in reverse select `handleBeta`. Same repository, same commit, a different
persisted graph — so a filesystem that enumerates differently, or an incremental
run that reorders assembly, silently changes what the tool reports.
Four sorts ranked by score or length alone and returned 0 on a tie.
`Array.prototype.sort` is stable, so a 0 preserves INPUT order, which traces
back to `graph.iterNodes()`. Under `maxProcesses` capping that decided which
`Process` and `STEP_IN_PROCESS` nodes were persisted at all. Each now falls
through to a totally-ordered, content-derived key — node id for entry points,
the joined path for traces.
WHAT IS ACTUALLY VERIFIED, stated precisely because "four fixes" would overclaim:
- the ENTRY-POINT sort is individually mutation-verified;
- the two DEDUP sorts are collectively mutation-verified;
- the TRACE-RANK tiebreak is NOT individually observable, and the source says
so. The dedup sorts already impose a total order on the list that reaches
it, so removing it alone fails nothing. Kept as defence in depth: it cannot
misbehave — it only makes an already-deterministic order explicit — and it
is what stops a change to dedup ordering from silently re-opening this.
Finding that out took two fixtures. The first (three chains, three entry points)
is separated by the entry-point sort before trace ranking is reached, so it never
exercises the trace comparator at all; the second gives ONE entry point two
equal-length branches to different terminals, which is the only shape where the
trace comparator decides. Both are kept — they gate different sites.
The invariance tests assert the INVARIANT rather than any single sort, so they
cover all four sites and any future one without needing to know where they are.
Three assertions: same selection under a cap, identical set uncapped, and
identical ORDER — the last because order is what the cap consumes, so a set-only
assertion would pass while the defect persisted.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(impact): UNKNOWN dominates a mixed candidate set, instead of reporting the known floor (W2-4)
Fourth wave-2 item. The all-UNKNOWN branch here was reasoned about carefully and
is correct — its comment even names the two ways a set can be all-UNKNOWN. The
MIXED case fell straight through it.
`RISK_ORDER` is `['LOW','MEDIUM','HIGH','CRITICAL']` and has no `UNKNOWN` entry,
so `indexOf('UNKNOWN')` is -1 and an UNKNOWN candidate can never win the reduce.
An ambiguous name with one caller-less candidate (UNKNOWN, per the round-1 fix)
beside one single-caller candidate (LOW) reported `maxRisk: 'LOW'` — a confident
floor over a set containing an interpretation nobody measured. That is the same
false-safe the all-UNKNOWN branch exists to prevent, one case over, and it
surfaced in the UI as "Max blast radius N (LOW risk)".
`maxRisk` answers "how bad could this be?", and an unresolved candidate could be
CRITICAL — so any UNKNOWN in the set makes the aggregate UNKNOWN. Narrowing it
that way would normally cost information, so the measured part travels alongside
as `knownMaxRisk`, present only when the two differ: absent on a fully-resolved
set, where it would duplicate `maxRisk`, and absent on a fully-unknown one, where
there is no measured part. A reader gets "at least LOW among what resolved, and
one interpretation could not be walked at all", which is strictly more than
either value alone. The human-readable message says the same thing.
The seed gained a mixed pair because the existing one could not reach this: both
its twins are caller-less, so it only ever exercises the all-UNKNOWN branch —
which is precisely why the gap survived a round of review. Three assertions,
both halves mutation-verified.
`eval-server.ts` needs no change: it renders `result.maxRisk ?? 'UNKNOWN'`, so it
now shows UNKNOWN where it previously showed the floor.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(routes): track ternary polarity in dispatch guards, so a selected verb cannot be inverted (W2-9)
`if ((req.method === 'GET' ? false : true) && pathname === '/api/i')` emitted
`GET /api/i` — the one method that branch guarantees the request does NOT have.
A ternary SELECTS between its arms, so a verb inside one is not reached merely
because the whole condition is truthy, but `findVerbInSubtree` descended into
both arms and returned the first verb it saw. Same inversion `!` produced before
|
||
|
DuduPhudu
|
223ac7010d
|
feat: close reported graph blind spots in reference resolution, analyze and storage (#2856)
* fix(mcp): report UNKNOWN risk when an upstream impact walk finds no callers `risk: LOW` asserts "safe to change" — a claim ABOUT callers. An upstream walk that resolved none has nothing to base it on: the symbol may be genuinely unused, or reached only through a reference class the index does not record (a property access on a plain object, a bare-identifier read of a module-scope const). Seeding LOW from an empty result is the false-safe signal `anyKnownRisk` already refuses to emit on the ambiguous-candidate path, and that #2687 removed by making an undetermined impactedCount `null` rather than `0`. Zero-caller upstream results now report risk UNKNOWN with a riskNote saying absence of edges is not evidence of disuse. Downstream is untouched: an empty downstream walk reports resolved callees, not safety. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(javascript): emit ACCESSES for bare-identifier reads of module-scope consts A constant read only as a bare identifier — `Math.max(LIMIT, n)`, a default parameter value, `return LIMIT` — minted no reference site at all, because JS captured only `@reference.read.member`, which requires a receiver a bare identifier does not have. So "who uses this constant?", the question behind every dead-code trim and constants refactor, answered with a confident zero in both directions. The rest of the machinery was already in place: `FIELD_KINDS` accepts `Const`, the scope query already declares it via `@declaration.const`, and `read` maps to ACCESSES for any resolved target. This adds the missing capture in VALUE POSITIONS ONLY (call arguments, default-parameter values, return statements) — a blanket `(identifier)` rule would mint a site for every token in the file, which is unaffordable at repo scale and would keep alive the block-local symbols `pruneLocalSymbols` exists to drop. Cross-file readers are NOT yet covered: the site exists and a call through the same import statement resolves, but a value-kind def does not link across the import edge. Recorded as a todo with the investigation. PARSE_CACHE_VERSION bumped 44 -> 45: this is parse-time capture emission, so a warm cache replays the pre-change capture set and the new edges never appear — observed directly, a full `analyze --force` produced a byte-identical graph until the cache was cleared by hand. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(javascript): pin A1/A5 plain-object property acceptance criteria Fixture plus todo specs for the four shapes plain-object property access has to answer: object-literal keys indexed as Property nodes, a read through the holding variable, a property WRITE, and a read through an untyped param. Records the investigation so the work is resumable: the parse-query pattern scoped to literals bound to a variable matches correctly (verified against the raw JAVASCRIPT_QUERIES), but no Property node reaches the graph and local-symbol-pruner is not the cause — it drops only Const/Variable/Static. The remaining gate is in the parse worker's node-creation path. No production code — specs only, so the suite stays green. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(javascript): index object-literal keys of a named object as Property nodes Idiomatic JS models configuration as an object literal, not a class, but Property definition nodes existed only for DECLARED CLASS FIELDS. A config field therefore had no symbol at all: `context({name: 'exitMinAtrMult'})` answered "not found" for a field read and written throughout a live code path, and ACCESSES had no target to point at. Both halves are added for keys of a literal BOUND TO A VARIABLE — the parse query mints the graph node, the scope query mints the def the resolver can aim at. Unbound literals are deliberately excluded: an inline call argument or a JSX prop bag is call-site data, not a named surface other code references, so a node per key there would add volume without adding an answerable question. This lands the definition-node half only. The ACCESSES edges still require receiver resolution — typing the const that holds the literal to the literal's scope for the precise case, and name-based matching at reduced confidence for the untyped-param (option bag) case. Both are recorded as todos with the mechanism each needs. Also records a trap that cost a wrong conclusion: under vitest the parse worker runs the BUILT dist code (parse-impl resolves parse-worker.js, absent under src/, and falls back to dist), so parse-query changes are invisible to tests until `npm run build`. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(cache): move the SCHEMA_BUMP pin to 45 The pin is the guard that makes two branches claiming one cache-schema number fail loudly instead of silently serving each other's entries, so a bump is only half-done until the pin moves with it. The bump itself landed with the JavaScript bare-identifier captures; this is the other half. Caught by the guard working exactly as designed — the suite failed with "expected 45 to be 44" rather than letting a mismatched pair through. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): resolve plain-object property access by unique name Idiomatic JS reads configuration off an object whose receiver cannot be typed — an options bag passed as a parameter, a destructured handle, an imported literal. No precise pass resolves those, so a field read and written across a live code path produced no ACCESSES edge at all and "who reads this setting?" answered a confident zero. A last-resort pass runs after every precise pass and sees only what they left behind. For each still-unresolved read/write site it asks whether exactly ONE Property in the workspace carries that name. If so the read almost certainly means it. If two or more do, nothing is emitted and the site is COUNTED as ambiguous — a guess between them would be a coin flip, and a wrong edge in the pre-edit safety gate is worse than a missing one. Uniqueness is the right gate because it recovers exactly the names worth recovering: distinctive domain fields (exitMinAtrMult, bookNotionalUsdt) are unique in a repo and resolve, while generic keys (id, name, data) are not and are skipped — which is where name matching would over-connect. Bounded four ways: - Confidence 0.5, the global tier, with the inference named in the reason, so a consumer can filter inferences without losing scope-resolved edges. - Never second-guesses a precise result: sites already resolved are excluded, because first-write-wins stops a duplicate but NOT a second edge to a different target. - Honors `fieldFallbackOnMethodLookup`. A statically-typed language opts out of name matching precisely because it over-connects; inferring an ACCESSES edge by name is the same claim and must obey the same opt-out. - Requires an explicit receiver — a bare identifier is not a property access, and matching one by name would link a local to an unrelated key. Indexes graph nodes rather than scope defs because an object-literal key mints a Property NODE but no scope-resolution DEF: `localDefs` and `scope.bindings` are both empty for exactly the population this serves. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(analyze): record a collapsed graph write instead of reporting fresh The dangerous half of a broken refresh: metadata IS written, so the index reads as fresh, hooks re-arm, and every tool answers from a graph missing most of its edges — indistinguishable from a codebase that genuinely has no such relationships. Reported in the field as edges collapsing 23009 -> 2170 and as a CodeRelation table that never materialized. `analyze` now compares the relationship count the pipeline PRODUCED against what the DB hands back after the write. Both numbers are already in scope at the same point, so the shortfall is provable rather than inferred — no comparison against the previous index, which cannot distinguish a failed write from a repo that legitimately shrank. A missing relation table needs no special case: it reads back as a persisted count of zero. On a collapse the run records `graphWriteCollapsed` in metadata, which `getIndexIncompleteReasons` turns into `graph-write-collapsed` so status and the MCP resources report the index INCOMPLETE rather than fresh. A ratio, not equality: some relationship types do not round-trip one-for-one and `--pdg` writes MORE rows into the same table, so demanding equality would fire on healthy runs. Only a collapse is a defect. Fail-safe when the expected count is unavailable — an implementation that offloads relationships out of memory may not be able to report a total, and a false "your index is broken" is worse than a missed one. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(ingestion): qualify object-literal Property ids by their owning object Two config objects in one file that share a key name generated the same `Property:<file>:<key>` id and COLLAPSED INTO ONE node, so two distinct settings became a single symbol. Worse, the merged name then looked workspace-unique to name inference, which happily resolved reads of it to a node representing both — a wrong edge in the pre-edit safety gate, which is precisely what the unique-name pass is bounded to avoid. `objectLiteralOwnerInfo` already existed for exactly this ("so two constructors in one file that both define `bar` stay distinct nodes") but was gated to `Method`. `Property` now opts in. `findObjectLiteralBindingInfo` returns `ownerName` only when asked. Its `Method` ids must stay byte-identical — qualifying them would rewrite every object-literal method id in every indexed repo — while object-literal KEYS, indexed only since A1/A5, have no such history to preserve. Found by a test written for the ambiguity path rather than by review: the suite reported one node where two were expected, and an edge where none should exist. Both are now pinned, along with the detection boundaries of the B2 collapse check, which was previously an untestable inline expression and is now a pure function. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(typescript): index type aliases and shape members as symbols A TS frontend models its API contracts as `type X = { … }` and `interface`, so a field on one is exactly what "who breaks if I remove this?" is asked about. Three gaps made that unanswerable, all in the TypeScript queries: 1. No `type_alias_declaration` -> `@definition.type`, so an alias minted NO NODE AT ALL and a context() lookup on an exported contract type answered "Symbol not found". TypeScript was the ONLY language missing this — Rust (type_item), Kotlin (type_alias), Swift (typealias_declaration) and Dart all emit it. The alias was declared for scope resolution but never became a graph symbol. 2. No `property_signature` in the parse query, so INTERFACE members minted no Property nodes either — the upstream report's "class/interface index fine" holds only for the type, not its fields. 3. No `property_signature` in the scope query, so even with nodes present the resolver had no member declaration to aim at. Its sibling `method_signature` -> `@declaration.method` already existed; only properties were missing. Interface bodies and object-type aliases both spell members as property_signature, so one pattern per query covers both shapes. Lands the SYMBOLS, not yet the ACCESSES edges: the shape is already a class-like scope and now has member declarations, but no edge forms — the remaining link is owner/type-binding, recorded as todos with the diagnosis. Note TypeScript sets fieldFallbackOnMethodLookup:false, so unlike JavaScript there is deliberately no name-based fallback here; the precise path is the only route by design. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(golden): accept interface members in the mini-repo snapshot Drift is entirely the new TypeScript shape-member indexing: the fixture's three interfaces (ValidationResult 2, DbRecord 3, LogEntry 3) contribute exactly 8 Property nodes, each with exactly one HAS_PROPERTY owner edge. Verified before regenerating rather than after: every pre-existing count is untouched (CALLS 9, IMPORTS 12, DEFINES 16, HAS_METHOD 1, MEMBER_OF 12, STEP_IN_PROCESS 12), so nothing was rewired — the digest moved only because 8 edges were added. The fixture's inline `return { valid: false, … }` literals correctly produced nothing, confirming the object-literal rule stays scoped to variable-bound literals. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(analyze): never report a collapse from a non-numeric count The B2 check reported healthy runs as total graph-write collapses. A non-numeric `expected` (a graph implementation reporting no total, a lightweight pipeline result) does not skip the guards — it INVERTS them: `undefined < 100` is false, so the small-repo exemption never fires, and `0 >= undefined * 0.5` is `0 >= NaN`, also false, so the ratio check "passes" as well. Both bounds silently evaporate and every such run is flagged. That is precisely the failure this check was written to catch, reproduced inside the check itself: an unmeasurable quantity treated as a measured zero. Both sides are now validated as finite numbers before any comparison. `persisted` is also passed as UNKNOWN rather than zero when the DB was not demonstrably readable: `getLbugStats` flattens "no connection", "query threw" and "empty table" all into `edges: 0`, so `stats.nodes > 0` is used as independent evidence the read happened at all. Caught by the existing run-analyze suites, not by the new unit tests — those exercised the pure function with well-formed numbers and were blind to the integration's actual inputs. Both cases are now pinned. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(typescript): make object-type aliases own their members A TS object-type alias declares the same `property_signature` members as the interface beside it and answers the same question, but was not a member owner: its fields were minted with bare ids and no owner edge, so two aliases in one file sharing a field name collapsed onto one node, while the identical interface resolved normally. `type_alias_declaration` joins CLASS_CONTAINER_TYPES (and CONTAINER_TYPE_TO_LABEL, as that set's invariant requires — a container missing there gets orphaned member edges or a wrong owner label). Aliases with no object type (`type Id = string`) declare no members, so they own nothing and are unaffected. This also lands the INTERFACE field -> consumer edges, verified on the mini-repo fixture rather than only on a purpose-built one: `saveToDb` now links to `ValidationResult.value`, and `formatLogEntry` to `LogEntry.level` and `LogEntry.message` — three real contract-field reads that previously had no graph path at all. Golden updated: +3 ACCESSES, no node changes. The ALIAS field -> consumer edge is still not linked and is recorded as a todo with the exact blocker: resolving a receiver typed as the alias needs the NAME to resolve to a class-like def, and `isClassLike` is Class|Interface|Struct|Record|Enum|Trait. That predicate is read from ~12 sites including MRO and heritage, and every language mints TypeAlias, so widening it would enrol aliases in linearizations where they do not belong. Widening only the scope index was tried and reverted — the type-name walkers gate on it independently, so it fixed nothing and left dead code. That needs a deliberate "shape-like" concept, not more call-site widening. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs(test): record the traced diagnosis for the unlinked alias field edge Traced to the end rather than left as "needs investigation", so the next attempt starts from facts: 1. Graph side is COMPLETE and symmetric with the interface — Property:...:LiveModeConfig.bookSlots is owner-qualified and carries HAS_PROPERTY. 2. Resolution DOES reach resolveClassBindingForName('LiveModeConfig') (instrumented) and misses. 3. It misses because the module scope binds LiveModeIface:Interface, renderAlias, renderIface — and not LiveModeConfig. The alias has no binding on the receiver's scope chain at all. 4. The TS scope query tags aliases @declaration.type, but normalizeNodeLabel accepts only typealias / type_alias and has no "type" case, so it returns undefined. Kotlin and Dart use @declaration.type_alias; TypeScript is alone on the dead tag. 5. Retagging is NECESSARY BUT NOT SUFFICIENT — tried, and the binding still does not appear, so a second gate exists in how a declaration anchored on a node that is ALSO a @scope.class anchor is attached: the alias appears to bind inside its own scope rather than hoisting to Module, where interface_declaration evidently does hoist. An isShapeLike predicate (the nominal-vs-structural split: shapes declare members, nominal types participate in MRO) plus a mirrored findShapeBindingInScope were built and REVERTED along with the retag. With no binding on the chain they never fire, and shipping inert widening is worse than shipping none — the same standard applied to the earlier scope-index attempt. The design is recorded here; it is worth doing once step 5 is fixed, and it also unblocks Rust's parked union_item, which the MEMBER_OWNER_NODE_TYPES comment documents as the same gap in another language. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): resolve cross-file value references, skip block-locals Two halves of the same question, "who uses this constant?". CROSS-FILE. `resolveReferenceSites` runs against the registries and, as its own comment says, "imports live in finalized bindings the registries can't see" — which is why free CALLS need `emitFreeCallFallback`. Reads had no counterpart, so `import { LIMIT }` followed by a bare use resolved to nothing while a CALL through the very same import statement resolved fine. This adds the read/write counterpart, reusing `findValueBindingInScope` (which walks the FINALIZED chain) rather than inventing a lookup. Confidence 0.9: the import names the def, so this is precise resolution, not inference. BLOCK-LOCALS. Bare-identifier capture also matches a read of a block-local `const`, and an edge to one keeps alive exactly the inert locals `pruneLocalSymbols` exists to drop — a pruned node becomes a retained node plus an edge, in every function of every indexed repo. Emission now takes the set of value defs bound at MODULE scope and drops ACCESSES to Const/Variable/Static outside it. The cross-file pass carries the same guarantee structurally: a def in another file cannot be a block-local of this one, so it skips same-file hits entirely. The block-local leak was already shipped in the intra-file A2 commit and was found only because a test was written for the guard rather than the feature — the same way the object-literal id collision surfaced. Verified on the full resolver matrix: 3172 tests, golden unchanged. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(lbug): diagnose a vanished staging CSV instead of surfacing a Binder error A forced rebuild could fail with "COPY failed for File: Binder exception: No file found that matches the pattern .gitnexus/csv/file.csv" and then an ENOENT on .gitnexus/csv/rel_Folder_File.csv — two engine-level messages that name neither a cause nor a remedy, which is where several field reports end. Only tables with rows > 0 enter the COPY manifest (csv-generator.ts), so an absent file was WRITTEN during this run and removed since. Both COPY loops now preflight and say exactly that, with the row count, both causes the reports point at (a second `gitnexus analyze` on the same repo — they share .gitnexus/csv — or an external cleanup of .gitnexus/), and the action to take. Scope note, deliberately narrow: this does not attempt to fix WAL corruption or checkpoint rotation. Those already have detection and recovery hints (isWalCorruptionError, WAL_RECOVERY_SUGGESTION, the configurable wal-checkpoint-threshold), and the ~6000 lines added to lbug/ + storage/ since v1.6.9 — index-lock.ts most of all, which serializes writers and plausibly closes the concurrent-run class outright — postdate every report in the window. Guessing at unreproducible durability faults would be speculation; making the one failure with NO handling legible is not. An existing overlap test induced this exact scenario (a manifest entry pointing at a missing csv) and asserted on the engine's wording. Its intent — that a node-COPY failure is rethrown at the FK barrier rather than swallowed — is unchanged and still asserted; only the message it matches moved. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): split shape-like from class-like, linking alias fields Completes A4: a field on a TypeScript object-type alias now links to the code that reads it, the last unanswerable half of "who breaks if I remove this?" for a TS frontend that models contracts as `type X = { … }`. `isClassLike` answered two questions that only coincide for classes: 1. does this declare MEMBERS I can look up? — a SHAPE (structural) 2. does this participate in inheritance / MRO? — a NOMINAL TYPE An object-type alias is (1) and emphatically not (2) — it has no supertypes and no place in a linearization. Widening `isClassLike` to buy (1) would have enrolled every language's aliases (Rust type_item, Kotlin/Swift/Dart typealias, C typedef) into MRO and heritage, so the two questions now get two predicates. Call sites split by which they ask, and their names already said which: `resolveInheritanceBaseInScope` and `resolveQualifiedInheritanceBase` keep `isClassLike`; receiver typing and member OWNERSHIP take `isShapeLike`. Three parts, each necessary and none sufficient alone: - `findShapeBindingInScope`, mirroring `findValueBindingInScope`'s established relationship to `findClassBindingInScope` (same walker, different accepted def-type), consulted only AFTER the class lookup misses so a class of the same name always wins. - `populateClassOwnedMembers` uses it, so alias members get an `ownerId` and are registered under the alias. Without this the receiver resolved to the alias and then found no members under it. - The TS scope query tags aliases `@declaration.type_alias`, not `@declaration.type`: `normalizeNodeLabel` accepts typealias / type_alias and has no "type" case, so the old tag mapped to NO label and TypeScript aliases produced no scope-resolution def at all. Kotlin and Dart already spelled it this way; TypeScript alone was on the dead tag. An earlier attempt concluded a further "scope-attachment gate" existed. That was wrong and is worth recording: scope extraction runs in the parse WORKER, which loads built `dist`, so the retag was never executed. Rebuilt, the alias hoists to Module scope exactly as the interface does. Same trap as the parse query — `src` edits to anything the worker runs are invisible until `npm run build`. Typedef and Union stay out of `isShapeLike` deliberately: they belong conceptually (the union_item note on MEMBER_OWNER_NODE_TYPES records the same gap) but neither is wired as a member container, so including them would widen a predicate nothing exercises. Verified on the full resolver matrix: 3173 tests, golden unchanged. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(typescript): pin the type-alias capture to a tag that maps to a label The capture test asserted `@declaration.type`, the tag that `normalizeNodeLabel` does not recognize (it accepts typealias / type_alias and has no "type" case). So the test passed for as long as the tag was broken: it checked only that the capture FIRED, never that it resolved to anything, while TypeScript aliases produced no scope-resolution def at all. Updated to the working tag and given a second assertion that the derived kind string is one the label mapper accepts — the property that actually matters, and the one whose absence let a dead tag sit pinned. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(lbug): declare TypeAlias member pairs so analyze does not abort Making object-type aliases member owners emits HAS_PROPERTY from a `TypeAlias`, and the relation schema declared no such pair. The emit therefore threw `UndeclaredRelationPairError` and the ENTIRE analyze died on any repo containing `type X = { ... }` — a hard stop, not a dropped edge. Found by running the analyzer over a real 16k-node TypeScript repo, not by a test. `Method` is declared alongside `Property`: a member written `type Handler = { onClick(): void }` is a method_signature and would fail in exactly the same way. Why every existing test missed it: the resolver suites build an in-memory graph via `runPipelineFromRepo` and never write to LadybugDB, so the schema constraint was never exercised. `structural-pair-coverage.test.ts` is the one suite that does run the emitters against the declared pairs — and its own docstring names the gap: coverage is bounded by NON_BRIDGE_CORPUS, "a new structural emitter should land with an entry here". This adds that entry, pinning TypeAlias|Property and Interface|Property as sentinels. Verified the guard is not vacuous: removing the pair again makes the suite fail with undeclaredPairs: ["TypeAlias|Property"]. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(processes): trace depth-first so multi-hop flows are detected D1 ("query ranks frontend components above the backend module that owns the concept") and D2 ("processes is dominated by trivial mechanical chains") are the same defect, and neither is about ranking or selection. The walk stops after a fixed NUMBER of traces, so traversal order decides which traces those are. Breadth-first reaches every shallow terminal before any deep one, so the quota filled with the shortest paths in the graph and the walk stopped — `maxTraceDepth: 10` was never approached. Measured on a real repo before the fix: of 300 processes NONE exceeded 7 steps and 90% were 3-4. A multi-hop business flow (signal → order → exit) therefore had no process that could represent it, and `query` could only rank the mechanical pairs that did exist. Step 4 of the caller already sorts by length and dedupes by endpoint — it was always asking for the deepest traces this walk could give it. Depth-first descends to a terminal first, so the same quota is spent on paths worth keeping. Cost is unchanged: same budget, same cycle guard, same depth ceiling — only the order differs. Measured on the same 16k-node repo, same build and flags, BFS vs DFS (an earlier comparison was discarded as confounded — it crossed builds and --pdg): steps 6-8: 50 → 168 (3.4x) totals: 844 → 806 and the reported query moved from `LiveSetupView → Cn` (a React component) to `ReconcilePositions → IsTpInProfit / WithHeld / ShouldNotify` — server-side exit management, which is what was asked for. `traceFromEntryPoint` is exported for the test. Traversal order is unobservable through `processProcesses`: `findEntryPoints` supplies several starting points, so a deep chain is traced from inside it whatever the order does. A test at that level passes under BOTH traversals — the first version of this test did exactly that and guarded nothing. Driving the walk directly, it fails under breadth-first with "expected 3 to be greater than 3". Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs(test): correct a stale status note left behind by a later fix The A1/A5 header still said "edge resolution REMAINING ... neither is implemented". Both shapes resolve — the typeable receiver precisely, the untyped one by workspace-unique name — and the tests below assert exactly that, so the note contradicted the file it sat on. It was accurate when written and went stale when the work continued past it. Left as-is it would tell a reviewer that a landed feature is missing. The TRAP note is kept: the parse worker still runs built dist under vitest, and that is still the trap it describes. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): index literals behind identity-preserving wrappers `export const INERT_EXIT_CONTRACT = Object.freeze({ ... })` minted no `Property` node for any of its keys. The object-literal rule matches `variable_declarator > value: (object)` as a DIRECT child, and freezing puts a call expression in between — so the shape whose fields are most worth querying was the one shape the rule could not see. Freezing a config object is how JS publishes an immutable contract, which is why this reads as a confident zero on exactly the fields a reader cares about. The allowlist is three functions, not "any call". `Object.freeze`, `seal` and `preventExtensions` RETURN THE ARGUMENT THEY WERE GIVEN, which is what makes the literal's keys members of the bound name. For `const x = compute({ a: 1 })` the literal is an argument and `x` holds compute's return value, so attributing `a` to `x` would be a fabrication. Two negative controls, because the obvious one is vacuous: a bare-identifier callee is rejected structurally and would pass with no allowlist at all, so the assertion that actually pins the predicate uses `Object.entries` — identical shape, differing only by name. Verified load-bearing by adding `entries` to the allowlist and watching that test alone fail. SCHEMA_BUMP 46 -> 47: parse-time emission, so a warm cache replays the pre-fix capture set. Observed as a false negative first — `analyze --force` returned the old node set until the on-disk cache was removed by hand. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(scope-resolution): narrow multi-candidate property names by scope Workspace uniqueness was the wrong denominator. Measured on the reporting repo: `exitMinAtrMult` has 26 `Property` definitions — 16 in one-off `scripts/`, 7 in the frontend, one in a test, and exactly ONE in the backend that reads it. Every backend read was refused because of competitors the reader cannot see. The gate was not too permissive or too strict, it was scope-blind. A name with several definitions is now narrowed before being abandoned: same-file first, then files the reading file directly imports, using the finalized import graph rather than a path-shape heuristic. Exactly one survivor at the first non-empty tier resolves; anything else stays refused. A tier holding several candidates stops the walk instead of falling through — local evidence that is itself ambiguous still contradicts reaching further out. Confidence stays 0.5 at every tier. Narrowing changes which candidate is chosen, not the kind of claim: it is still a name match, and the round-1 contract is that filtering on confidence drops all name inference at once. The reason string now names the tier that fired. Ambiguity reporting goes from a count to the actual names (capped), because a count says a gap exists while the names say which fields are unanswerable. Measured on that repo, backend readers of `exitMinAtrMult` go 0 -> 24 and total readers 9 -> 45, including the two call sites in `oppositeSignalExitManager.js` the report singled out. Both narrowing tests were mutation-checked by dropping the import evidence and confirming they, and only they, fail. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): capture destructured parameter keys as property reads `function exit({ exitMinAtrMult = 0 })` reads that property off whatever the caller passes, exactly as `cfg.exitMinAtrMult` would. It never appears in a member_expression, so it had no reference site at all — and this is the shape the function that IMPLEMENTS a behaviour uses, so the most relevant reader was the one systematically missing from "who reads this setting?". Uses a distinct `@reference.read.destructured` anchor rather than `@reference.read.member`. The latter is filtered emit-side to matches with a member_expression ancestor, because calls and writes share its shape, and a destructuring pattern has none — reusing the tag would have been silently dropped by that filter. The `read.` head already maps to a read kind, so no mapping change is needed. Scoped to formal_parameters. A destructuring binding elsewhere (`const { x } = require('m')`) is frequently an import rather than a field read, and minting a property read there would attribute module bindings to unrelated same-named keys. All three cases (default value, bare shorthand, renamed key) mutation-checked by removing the patterns and confirming those three tests, and only those, fail. The renamed case also asserts the edge points at the KEY and that the local alias mints nothing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(scope-resolution): link type consumers to the type they name An exported contract type owned its members after round 1 and still answered `incoming: {}`, so "what breaks if I remove this field?" — the question a contract type exists to answer — had no edge to walk. Measured on the reporting repo: all 324 TypeAlias nodes AND every Interface node had DEFINES as their only incoming edge. Two independent causes, and the second is why the first was not enough. TypeScript captured no type references at all — only cpp and csharp did — so an annotation naming a declared type minted no reference site. Added for annotations, generic arguments and `as` assertions, anchored to those contexts rather than a bare `(type_identifier)`, which would also match the name in `type X = …` and make every declaration a consumer of itself. That alone fixed interfaces and left aliases still empty. `TypeAlias` was missing from `LINKABLE_LABELS`, so alias graph nodes were never indexed in `nodeLookup` and `resolveDefGraphId` could not bridge a def to its node — the edge was dropped AFTER a successful lookup. `CLASS_KINDS` has always listed TypeAlias and the ClassRegistry returned the def correctly, which is what made this read as a resolution failure; instrumenting the lookup showed it returning the right def all along and moved the search one table over. Exactly the bug already documented two entries above it for Trait. Fixes every language that spells an alias this way — TypeScript, Kotlin, Dart and Rust all emit `@declaration.type_alias`. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): capture record construction as property writes The read side answered well after the narrowing work while "who SETS this field?" still missed the code that stamps the value. A record built inline — `return { exitContract: { exitMinAtrMult: settings.x } }` — is bound to no variable, so it minted no definition and its keys referenced nothing. Modelled as WRITE REFERENCES, deliberately not definitions. The round-1 rule already mints Property nodes for literals bound to a variable; minting more for anonymous records would add same-named competitors to the very name-narrowing that makes these fields resolvable — measured at 26 competing definitions for one field on the reporting repo, which is what made every backend read unanswerable in the first place. A construction site is a USE of a field, not another declaration of it. Two positions only: nested under a key, and returned. Both are records with a name attached (the key, or the function). An inline call argument (`doThing({ id: 1 })`) stays excluded for the same reason round 1 excluded it from definitions — it is call-site data, not a named surface — and is asserted as such. The enclosing literal is the receiver and it is anonymous, so these route through the same narrowing and the same refusal-to-guess as every other untyped receiver. Verified on the reporting repo: `entryPlan.js` went from no rows to `selectExitEnvelope` as a writer of `exitMinAtrMult`. Both captures mutation-checked. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(processes): select round-robin by terminal so the list is not one flow repeated Ranking was `sort by length` alone, so the top of the list was one behaviour described many ways: eleven of the top fourteen processes on the reporting repo were four entry points crossed with three terminals of the SAME date-window utility cluster. Genuine call chains, but a reader learns one thing from fourteen entries, and the repo's own domain flows sat below them. Selection now round-robins across TERMINALS, deepest first. Depth still orders within a terminal and still leads the list; what changes is that no terminal takes a second slot until every other has had a first. Keying on the entry point was tried first and made it worse — many files declare a `main`, so each was a distinct entry that round-robin then awarded its own slot, and `Main -> AlignWindowEnd` went from one row to eight. The repetition was never in where a flow starts. Measured on that repo: distinct terminals in the top 20 went 3 -> 20, and its domain flows (`ReconcilePositions -> ...`) moved into the top 4%. Two things this deliberately does not claim. The reported cause — ranking rewarding fan-in, promoting chains ending in widely-called helpers — measured FALSE: those terminals have one caller each (`alignWindowStart` 1, `validateSymbol` 1). A fan-in discount was implemented against that hypothesis, measured, and reverted for moving nothing. And a business flow still cannot be a process in its own right: the walk only emits at a leaf, at max depth, or on a cycle, so a flow whose meaningful endpoint calls onward survives only as whatever leaf it bottoms out in. Both are recorded in the code so neither reads as settled. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test(structural-pairs): pin the type-annotation USES pair R2-2 emits USES INTO a `TypeAlias`, so the pair is `Function|TypeAlias` — a different table from the `TypeAlias|Property` entry added in round 1, and one that entry stays green without. `TypeAlias` is on the eleven-table list this suite exists for, and an undeclared pair does not degrade: it throws `UndeclaredRelationPairError` and kills the entire analyze on any repo containing an annotated type. Every resolver suite still passes, because they build an in-memory graph and never write to the DB. That exact failure shipped once in this PR already. Two emitters into the same label, each with its own way to reach a released build. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(scope-resolution): build the module-level set before the out-of-core seal Review blocker. Under `GITNEXUS_DISK_SCOPE_INDEX=1` the seal replaces every ParsedFile with a scope-STRIPPED copy, and the block-local filter's set was built after it — so it walked `scopes: []` for every file, came out empty, and the filter read that as "no def is module-level" and dropped EVERY `Const`/`Variable`/`Static` ACCESSES edge in the repo. All languages, all files, including the module-scope-const edges this PR exists to add. Nothing threw and nothing logged, on the path the largest repos take: the exact confident-empty answer the PR is about. Built above the seal now, from `parsedFiles`, and passed as `undefined` rather than an empty set when no scope was inspectable — an empty set is a legitimate answer ("this repo has no module-level value defs") and must not be indistinguishable from "could not look". Fails open; the block-local exclusion is still asserted under the seal, since that is correctness rather than optimization. Also widens module level past `kind === 'Module'`. A `Namespace` scope (TS `namespace`, Rust `mod`, C++/C# `namespace`) holds importable values too, and treating its consts as function-locals dropped their reads. Included only when the whole chain to the root is Module/Namespace, so a namespace declared inside a function body stays local — asserted both ways. That fixture then failed for a third reason: `@reference.read.identifier` existed ONLY in the JavaScript query, so A2 did not work for TypeScript at all. Added there, and both languages widened to `variable_declarator value:` and `binary_expression` operands — the gaps review named between what A2 claimed and what it matched. Nothing covered `GITNEXUS_DISK_SCOPE_INDEX`. The new parity test asserts the seal changes no edge, and was verified against an emulation of the original bug: same-file readers vanish and only the cross-file reader survives. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(typescript): anchor property_signature to declared shapes Review blocker, and it reproduces end to end. `property_signature` occurs in EVERY object_type in the TS grammar, not only in an interface body or an alias's object type, so inline parameter types, inline return types and nested object types all matched — and the enclosing-container walk hung each one off the nearest class, interface or alias. Measured against the unanchored rule, all four appeared as members of shapes that do not have them: Property:contracts.ts:Svc.inlineParamOnlyKey Property:contracts.ts:Repo.inlineQueryOnlyKey Property:contracts.ts:NestedConfig.nestedOnlyKey Property:contracts.ts:buildInline.inlineReturnOnlyKey@46:33 When the inline member shares a name with a real one — `run(opts: { retries: number })` inside a class that declares `retries` — `addNode` is first-write-wins and the two distinct symbols merge onto one node, so every context()/impact()/rename() answer about that field describes the merge. The sibling JS object-literal rule in this same PR is anchored for exactly this reason; this is the TypeScript half of the same fix. `(A (B))` matches DIRECT children, so nested object types are excluded by the same anchor rather than by a second rule. The first version of these tests was VACUOUS and is recorded here because the reason generalizes: a collision and a correct exclusion both leave exactly one node behind, so counting ids cannot distinguish them. Every inline member in the fixture is now uniquely named, which is the only thing that discriminates — verified by restoring the unanchored rule and watching exactly those four assertions fail. A fifth test asserts anchoring costs no real member. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(analyze): correct the numbers feeding the graph-write-collapse guard Review blocker. The predicate itself held under adversarial probing; every defect was in what it was handed and what happened after it fired. (a) `expected` was wrong twice. Under `GraphEmitSink` streaming the bulk types leave the heap at parse time and never enter `relationshipCount`, so the count understated the real volume by most of it and the ratio passed trivially — on `force === true` runs, which include crash recovery AND the `analyze --force` retry this check's own warning tells the operator to run. Adds the manifest totals, the same correction the buffer-pool hint in this file already makes for the same reason. Separately, an incremental run persists only the changed subgraph while both counts are whole-scope: a 10,000-edge index that lost 200 replacements reads 9,800 and is certified complete. The check is skipped on that path rather than answered wrongly. (b) A throwing edge count became a measured zero. `getLbugStats` initialised its total to 0 and ran the query in a swallowing catch, so WAL/lock contention during finalize — documented on this exact call — reported a healthy index as a total collapse. It now returns `number | undefined`, and the caller requires both a readable node count and a defined edge count. (c) A total loss was exempted for being small. The min-edges rule tested `expected` before looking at `persisted` at all, so `expected = 99, persisted = 0` — every edge gone — stayed fresh and reported success. Total loss is now decided first. The existing test asserted the defect; it now asserts a PARTIAL shortfall, which is the case the exemption was written for. (d) A detected collapse reported success and exited 0. It is different in kind from the other incomplete reasons: those describe a run that did what it said and left work for later, this one means most of your edges are gone and every query answers a confident empty. The CLI now prints INCOMPLETE with the counts and sets a non-zero exit code, and the flag crosses IPC so the worker cannot send a clean `complete` either. Nothing exercised this wiring — only the pure helper. Adds tests for all four, each written so the pre-fix arithmetic fails it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(scope-resolution): keep unique-name property inference inside one language The pass indexed `Property` nodes from the whole shared graph. Per-language gating decides whether it RUNS for a language; it never restricted which nodes could be TARGETS. So the only carrier of a name could be in another language entirely, and a read here resolved to it on name uniqueness alone — no owner, no file, no call path. Reproduced: a Java class declaring `private int loyaltyPointsBalance` and a JS `cfg.loyaltyPointsBalance` on an untyped parameter produced an ACCESSES edge from the JS function to the Java private field. Confidence does not mitigate it, because `minConfidence` defaults to 0 — the tier is only a filter for consumers who ask for one. Candidates are now restricted to files in the language's own `parsedFiles`, which is a precise restriction rather than a heuristic and needs no new node property. Every other fixture in the suite is single-language, so this could not be caught anywhere by construction. The new fixture is deliberately polyglot and asserts both halves: no cross-language edge, and a same-language unique name still resolves. Known and not addressed here: the index is still O(total graph nodes) and is rebuilt once per qualifying language, the per-language whole-graph-scan pattern `phase.ts` hoisted out for `sharedNodeLookup`. Hoisting it belongs with that machinery rather than in this fix. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(processes): explore siblings in source order, log the exhausted budget `slice(0, maxBranching)` selected the FIRST N callees while `pop()` explored them LAST-first, so the trace budget went to the last-declared branch. For `main() { init(); loadConfig(); run(); shutdown(); }` the walk spends itself on `shutdown` and can drop `init` — the earliest steps of a flow, which is the opposite of what a process describes. Selecting first-N and exploring last-first was simply inconsistent; pushing in reverse makes the stack pop in source order. Measured on the reporting repo, this costs depth: 6-8 step processes go 168 -> 146 of 816. Still roughly three times the pre-PR baseline of 50, and the right trade — a deep branch is no longer reached by accident of being declared last. The remaining limit is the BUDGET, not the traversal: with a fixed quota a deep branch declared after enough shallow ones is not reached at all. That is now asserted in both directions rather than left implicit, and the walk logs when it stops with branches unexplored — a silently truncating cap reads as "this is everything", the same confident-empty answer this work is about, and the repo already sets that precedent for `dispatchFanoutSkipped`. Removes the second depth test, which was vacuous: the note twelve lines above it already said a `processProcesses`-level depth assertion passes under BOTH traversals, and measured it does — breadth-first yields the same deepest stepCount of 8, so it passed with the production change reverted. Traversal order is asserted against `traceFromEntryPoint` directly; what is observable at the pipeline level is which traces survive selection, which the diversity tests cover. Also renames `queue` to `stack` and corrects the BFS references in the module docstring and the function's own JSDoc, which is what an IDE hover shows. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(impact): carry riskNote onto ambiguous candidates and separate UNKNOWN's two meanings Two problems on the ambiguous fan-out, which builds its own candidate object rather than returning the single-symbol shape. The narrowed type had no `riskNote` field and never read one, so a candidate that resolved and found no callers reported `risk: UNKNOWN` with nothing attached — losing the entire point of the change on the path where the reader has the least context, since the name is ambiguous there by definition. And `UNKNOWN` used to mean exactly one thing on this path: the probe threw. The zero-caller branch gives it a second meaning, so an all-UNKNOWN fan-out could no longer be told apart from a broken one. Candidates now carry `probeFailed`, and the comment asserting the old reading is corrected. Also aligns `gitnexus-web`, which review flagged as giving a different verdict for the same symbol. That surface answers in prose rather than an enum, and its message said the symbol "appears to be unused (not called by anything)" — the identical false certainty in words. It now carries the same MEANING rather than the same field. Downstream wording is unchanged: no outgoing dependencies really is a fact about the symbol itself. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * test: replace assertions that cannot fail Four from review, each satisfied by the defect it was meant to catch. `new Set(props).size === 2` over two different literal strings can only ever be 2, so it could not detect the node merge its title promises — that is a difference in COUNT, now asserted on the raw array. The ambiguity test asserted only an empty edge set, which is satisfied equally by "the gate fired" and "the name was never looked up". It now also requires the ambiguity counter to have moved. `Interface|Property` was listed as a structural-pair sentinel beside `TypeAlias|Property`, but both its labels are in the SCOPE_BRIDGE cross-product so the pair is generated by construction and the sentinel cannot fail. Dropped rather than left reading as coverage; `TypeAlias|Property` is the load-bearing one. `TypeAlias|Method` was declared in the schema with no fixture emitting it — a declared pair no emitter exercises is indistinguishable from a missing one until an analyze aborts on a real repo. Adds a method-shaped alias member, and the suite requires sentinels to actually appear, so it is not vacuous. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * docs: document the new incomplete reason, the UNKNOWN verdict and the id churn Review found the code changes landed without the guidance around them, and an agent following this repo's own rules would have been told the wrong thing. `graph-write-collapsed` joined `INDEX_INCOMPLETE_REASONS` with no Sign block and no recovery section, while the precedent it cites (`embedding-checkpoint-pending`) has both — so `gitnexus status` would surface a new string naming silent wrong answers with nothing explaining trigger or remedy. Added to RUNBOOK and GUARDRAILS, including why this reason alone also fails the exit code. `AGENTS.md` said MUST warn on HIGH or CRITICAL and never mentioned UNKNOWN, and the shipped impact skill's risk table had no UNKNOWN row and still implied few-callers ⇒ LOW. An agent obeying those rules literally sees `risk: UNKNOWN` and proceeds, which negates the change the verdict exists to make. Both copies of both skills updated. `MIGRATION.md` now records that process ids do not survive this release — positional ids plus depth-first tracing, source-order siblings and round-robin selection mean `proc_7_handle` is a different flow afterwards. Bounded honestly: nothing in-repo joins on a raw process id, so it is index churn, not a broken consumer. `ARCHITECTURE.md`'s scope-resolution stage list gains the two new stages. The guide skill's node list gains `Property` and `TypeAlias` — the two node types this work most prominently creates. Also, on the pair-CSV preflight review asked to confirm: the hard abort IS deliberate, because a fallback recovering zero rows is the confident-empty failure this work targets. But the transient the message itself names — a second concurrent analyze sharing `.gitnexus/csv` — is a race, so the check now re-looks three times over ~150ms before declaring the file gone. Long enough to ride out a rename, far too short to mask a file that is genuinely missing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: drop redundant TypeAlias pairs and keep bare identifiers off class members Two regressions the full suite caught after the review fixes, both real. `schema-pair-coverage` failed with eleven hand-declared pairs that a rule now generates. Adding `TypeAlias` to `LINKABLE_LABELS` — needed so `resolveDefGraphId` can bridge an alias def to its node — also makes it a SCOPE_BRIDGE source and target, so the cross-product produces `File|TypeAlias`, `TypeAlias|Property` and nine others that round 1 had declared by hand. Removed; the invariant is that no pair is both generated and hand-declared. This also changes what the structural-pair sentinel means, and the comment is corrected rather than left overstating it: `TypeAlias|Property` is no longer load-bearing because the label is off the generated grid — it is load-bearing because it now depends on `TypeAlias` being IN `LINKABLE_LABELS`. Remove it and the pair stops being generated while the hand declaration is gone, which is the same state that silently breaks alias consumer edges. `block-scope-shadowing` failed because a bare identifier resolved to a class `Property`. `class Box { baseUrl = '...'; pick() { const baseUrl = ...; return baseUrl; } }` linked the block-local read to `Box.baseUrl`, duplicating the legitimate `this.baseUrl` edge. A bare identifier is not a member access: with no receiver there is no object whose property it could be, and in JS/TS a field read needs `this.`. Receiver-less read/write sites no longer accept `Property` hits; callables stay reachable, so `cb = save` naming a top-level function is unaffected. That defect PREDATES this branch's TypeScript captures — JavaScript has emitted bare-identifier reads since A2 and no class fixture exercised the shadow. The TS parity added here is what surfaced it. Golden snapshot regenerated after verifying the drift line by line: exactly +5 USES from type annotations in the mini-repo, every pre-existing count unchanged, so nothing was rewired. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * perf(scope-resolution): share the property-name index across language passes Review follow-up. `indexPropertyNodesByName` scanned every node in the graph and was rebuilt inside each qualifying language pass, reintroducing exactly the pattern `phase.ts` hoisted out for `sharedNodeLookup` — whose comment records why it matters: "the previous per-language rebuild burned that CPU+heap N times and, on a huge repo, a tiny language's full-graph copy overlapped the next language's — a real contributor to the scope-resolution memory peak." Built once in `phase.ts` beside `sharedNodeLookup` and `sharedFnNodeIndex`, and threaded through the same `prebuilt*` seam, so tests and isolated calls still build their own. Sharing is only safe because the per-language restriction MOVED rather than disappeared: the shared index is whole-graph, and candidates are filtered to the language's own files at lookup time. That also fixes a subtlety the per-language build had backwards — the cap now applies to the FILTERED set, so a name carried by forty properties across a polyglot monorepo but only two in the language being resolved is still answerable, where a global cap would have refused it. The tri-state at the lookup boundary is deliberate and the three outcomes are not interchangeable: no property of this name in this language (nothing to say, and NOT an ambiguity), too many to choose between (reportable), or a list to narrow. Caught mid-change by the polyglot fixture: an intermediate state shared the index without moving the filter, and the cross-language edge came straight back. That test earning its keep twice is the reason it exists. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): report when a field's only anchor is another language Round 3, found OUT-OF-SAMPLE — six field names appearing in no prior report, so nothing here was tuned against them. All six answered 0 backend ACCESSES while their definitions sat in `apps/research-dashboard/**`: TypeScript only. The in-sample set scored 5/5 and the out-of-sample set 0/6, and the gap is entirely this. Per-language inference (`3c5eadc7`) is right and stays. What was wrong is that declining is INVISIBLE: an empty result for a field anchored only in TypeScript is byte-identical to an empty result for a field nobody reads. One says "look in the other language or grep"; the other says "delete it". That is the same confident-empty failure this series exists to remove, one surface over — and this time the missing fact is about the ANALYZER's reach rather than the code. Declines are now counted and named, with the languages the anchors actually live in, kept SEPARATE from ambiguity because the remedies differ: ambiguity wants better receiver typing, this wants an anchor in the reading language. Collapsing them would tell a reader the wrong thing to do. A non-zero count warns at analyze time regardless of dev mode. The facts are published as `PipelineResult.propertyInference`, which they had to be for any of this to be testable — and that exposed a second defect. The round-2 ambiguity assertion, which I told the reviewer of #2856 I had strengthened, read its stat off a `scopeResolution` field that does not exist on PipelineResult: the `if (undefined) return` guard swallowed it and the test passed with the production code deleted. Both that assertion and the new ones now read the published field, and the guard is an assertion rather than an escape. Verified by deleting the counter and watching them fail. Reported by the same round-3 method note that caught it: verifying a fix against the cases it was written for only proves those cases pass. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(context): explain an empty property result caused by a cross-language anchor The other half of R3-1. The analyze pass now knows which fields it declined to link because every definition of the name lives in another language; this puts that fact where it is actually read. `context()` on such a field previously returned an incoming list byte-identical to a genuinely unread field. The two demand opposite actions — "look in the other language, or grep" versus "delete it" — so the difference has to travel with the answer: unresolved: property reads of this name were NOT linked: every definition of it is typescript, and name inference does not cross languages. An empty or short incoming list here is not evidence the field is unused — confirm with a text search, or give it an anchor in the reading language. anchorLanguages: ['typescript'] Carried through repo meta because the graph cannot answer it: the unlinked reads mint no edge and no node, so the only record is the pass that declined them. Keyed on the NAME, not on the resolved label. Gating on `=== 'Property'` was tried first and is wrong — the label reads `''` on this path for a plain Property node, so the gate silently suppressed the entire feature while every test still passed. Caught by asserting the field is DEFINED rather than guarding on it, which is the same anti-pattern that made two earlier assertions vacuous. The meta list only ever contains property names, so matching the name is itself the type check. Cached per (index, indexedAt): `ensureInitialized` deliberately avoids a per-call `loadMeta` because every tool routes through it, so this re-reads exactly when a re-analyze could have changed the answer and never otherwise. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): report declined property reads for opt-out languages too Generalizing R3-1 rather than waiting for it to be re-reported in the other direction. The reported case was a JavaScript read whose only anchor was TypeScript; the mirror — a TypeScript read anchored only in JavaScript — was still silent, because a language that sets `fieldFallbackOnMethodLookup: false` had the whole pass skipped, and skipping emission also skipped REPORTING. Detection is not inference. Counting what could not be linked asserts nothing about what it means, so `reportOnly` runs the pass for its facts while emitting no edge, and the opt-out keeps protecting exactly what it protected before. Two things this turned up that a single-instance fix would have missed: The cross-language fixture could NOT prove `reportOnly` is load-bearing — the per-language candidate filter already blocks those edges, so the assertion passed with the flag forced off. The case that discriminates is a SAME-language TypeScript read that name inference could legitimately link and the opt-out forbids; forcing the flag off there emits `readsTsOnly -> tsOnlyBudget`, which is the violation. Getting to that case surfaced a sibling gap, recorded but NOT fixed here: the object-literal `Property` rule is JavaScript-only, so `const CONFIG = { ... }` in a `.ts` file mints no node and its keys are invisible. The first draft of this fixture used exactly that shape and could not discriminate for that reason. It is the TypeScript half of R2-1a and wants its own change, not a rider on this one. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(typescript): index object-literal keys, as JavaScript already did The sibling recorded in `0c5a4f64` and deliberately left out of it. Both the named object-literal rule (A1/A5) and the identity-wrapper rule (R2-1a) lived only in JAVASCRIPT_QUERIES, so the single most common config idiom in TypeScript — export const tsRuntimeConfig = { tsConfigRetries: 3 }; — minted no node for any key. `context()` answered "Symbol not found" and a precise read through the holding variable had nothing to resolve to. TypeScript sets `fieldFallbackOnMethodLookup: false`, so these gain no name-based inference. What they gain is the PRECISE path, which is the route TypeScript is meant to use: `tsRuntimeConfig.tsConfigRetries` has a typeable receiver and now resolves. A read through an untyped receiver stays unresolved and, since `0c5a4f64`, is reported as such rather than answering an empty set. Scoped exactly as the JavaScript rules are — bound to a variable, and for the wrapper only the three functions that return the argument they were given — with the same `Object.entries` negative control pinning the allowlist. Found by fixture, not by report: the first draft of the `reportOnly` test used a TS `const CONFIG = { ... }` as its discriminator and could not discriminate, because the shape mints nothing. That is the whole argument for sweeping a class instead of waiting for each instance to be filed. SCHEMA_BUMP 47 -> 48: parse-time, so a warm cache replays ParsedFiles carrying none of these matches and the keys stay invisible. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): anchor anonymous returned object literals to their function The last gap round 3 named, and the dominant shape in idiomatic JS: 437 `return {` sites in a single backend directory of the reporting repo, including the ~25-field payload of its entire signal pipeline. The literal binds to nothing, so its keys could not even be named — "who reads wickRatio?" had no symbol to ask about. The enclosing FUNCTION is the owner: the literal is that function's return shape, a contract its callers consume. Keys qualify as `<function>.<key>`, so two functions returning the same name stay two shapes rather than one merged symbol, and multiple returns in one function stay distinct by position. RECONCILING THIS WITH R2-1b, which deliberately modelled returned keys as WRITES to avoid adding same-named competitors to narrowing. These are definitions, but narrowing now ranks DECLARED anchors — named literals, class fields, interface and alias members — strictly above return shapes. A name that already resolved keeps resolving to what it resolved to before, so the competitor problem R2-1b was avoiding cannot come back. Mutation-checked: dropping that ranking breaks five pre-existing R2 resolutions. That also required an R2-1b assertion to change, and the change is a strengthening rather than a concession. It asserted `toHaveLength(1)` — no new definition — as a proxy for "adding definitions must not move an existing answer". The proxy is now false while the property still holds, so the property itself is asserted directly. No `HAS_PROPERTY` edge from the function: that would be a `Function|Property` relation pair the schema does not declare, and an undeclared pair does not degrade — it throws and kills the whole analyze. That already shipped once in this PR. Two things found by dumping rather than assuming, both fixed here: SHORTHAND keys were not matched at all. `return { symbol, interval, score }` is the commonest spelling and the reporting repo's own payload is mostly this form, but tree-sitter models it as `shorthand_property_identifier`, which `(pair)` does not match. Caught by dumping the golden fixture and seeing a literal returning `{ level, message, timestamp: Date.now() }` had indexed only `timestamp`. Now covered in return position AND in the variable-bound rule, which had the same gap. Provenance was flagged by owner-presence, which mislabelled the anonymous case: a callback's return shape yields no name to qualify by, so it looked like a DECLARED anchor and would have outranked real declarations. Flagged by position now — a different question from whether a name could be derived. SCHEMA_BUMP 48 -> 49. Within one PR the version only has to differ from main's, but a build stamped 48 was installed and used to analyze before these captures existed, so caches stamped 48 carry none of them — the intermediate-build hazard this ledger already records for 33/34. Golden regenerated after verifying the drift: exactly +10 Property and +10 DEFINES, every pre-existing count unchanged. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix(scope-resolution): rank production anchors above test fixtures Found by testing R3-4 on the reporting repo instead of on its fixtures. Anchoring returned literals took `wickRatio` from 6 definitions to 13 — and backend reads still resolved to nothing, because SEVEN of the new JavaScript anchors compete and four of them are in `tests/`. A test constructs throwaway shapes carrying production field names; a read in shipped code cannot mean one of them. Applied before the declared/return-shape split, because "is this the shipped program" is the stronger signal — a declaration inside a test fixture is still a test fixture. Skipped when the READER is itself a test, since a read there legitimately means the test's own shape. The first version of this test was vacuous and the mutation check caught it: the reader sat in the same file as the production anchor, so the same-file tier resolved it whether or not this tier existed. The reader now lives in a file that imports neither anchor, which leaves production-vs-test as the only thing that can decide. Honest about what this does NOT do: it narrows `wickRatio` from seven candidates to three, and three functions in different files each returning that field is GENUINELY ambiguous — refusing is correct, and the ambiguity is now counted and named rather than silent. The reported question ("who reads wickRatio?") is answerable only where one producer exists; where several do, the honest answer is the list of producers, which R3-4 made nameable for the first time. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * feat(scope-resolution): resolve members through a call result's return shape The question three rounds of reports could not answer, and the one narrowing must refuse by design: a field produced by SEVERAL functions. A read of `spike.wickRatio` could mean any producer, so name inference correctly declines and no amount of tier-tuning changes that. It needs evidence, not inference. The evidence existed in two halves that had never been joined. The call-result type binding (`const alert = formatSpikeAlert(row)` binds `alert` to a TypeRef whose rawName is the callee) predates all of this work; it simply had nothing to resolve to when the callee returned an anonymous literal, because an anonymous literal named nothing. R3-4 gave it a name. Joining them: const alert = formatSpikeAlert(row); alert.wickRatio -> Property:...:formatSpikeAlert.wickRatio Precise, at ordinary emission confidence, and it works EXACTLY where narrowing cannot: several producers sharing a field name stop being competitors because the receiver says which one. Runs before the name fallback and claims its sites, so a precise answer is never second-guessed by a name match. Measured on the reporting repo: 1,410 precise edges, and all six fields round 3 verified OUT-OF-SAMPLE go from 0 backend readers to 7, 11, 10, 7, 6 and 14. Round 3 scored 0/6 on that set; this is 6/6. The bound is asserted, not just documented: a read off a BARE PARAMETER has no binding here, because typing it needs the caller's type to flow in — that is inter-procedural and genuinely larger. Those reads still fall through to name inference and are still reported when it declines. The fixture has two producers sharing a field name precisely so the test cannot pass by name matching, and mutation-checking the owner lookup fails it. No SCHEMA_BUMP: this is scope resolution, not parse-time capture, so a warm cache already carries everything it reads. Noted in the ledger because the reflex on this branch has been to bump, and an unnecessary bump costs every user a full re-parse. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Revert "return-shape anchoring" (R3-4/R3-5): it degrades query Reverts |
||
|
drdave
|
021ac30376
|
feat(cli): add a bunx lane so bun-only machines can run gitnexus (#2765)
Some checks failed
CodeQL / Analyze (javascript-typescript) (push) Waiting to run
CodeQL / Analyze (python) (push) Waiting to run
Gitleaks / gitleaks (push) Waiting to run
Publish / Classify release event (push) Waiting to run
Publish / RC guard (marker + release-PR skip) (push) Blocked by required conditions
Publish / ci (push) Blocked by required conditions
Publish / Publish to npm (push) Blocked by required conditions
Publish / Build & Push RC Docker images (push) Blocked by required conditions
Scorecard / Scorecard analysis (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-cli) (push) Waiting to run
Trivy Image Scan / Trivy (gitnexus-web) (push) Waiting to run
Skill copy sync / shipped skills drift guard (push) Has been cancelled
* feat(cli): add a bunx lane to the runner ladder
The ladder assumed a Node toolchain: global gitnexus, then pnpm dlx or
npx in some order, with npx as the last resort. On a bun-only machine
npm, npx and pnpm are all absent, so every rung fell through to npx and
both the emitted hint and the generated .gitnexus/run.cjs produced a
command the machine could not run at all.
Add bun as a fourth mode, invoked as an install-free bunx one-shot, on
two rungs:
- npm 11+ with no pnpm to fall back on — bunx dodges the same arborist
install crash the pnpm rung exists for (#1939);
- npm and pnpm both absent — previously the dead end described above.
Every pre-existing outcome is preserved: pnpm still wins on npm 11+, npx
still wins on npm < 11, and pnpm still wins over bunx when npm is absent.
Regression tests pin each of those. The bun PATH probe is lazy, so a
machine with a Node toolchain pays no extra scan and the stale-index hook
budget is unchanged.
bunx takes no allow-build equivalent: bun's --trust is a bun add/install
flag that writes trustedDependencies into a project package.json, which a
one-shot has none of, so the argv stays flag-free.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016R9psS9gJ73MRyquoBoPKg
* fix(lbug): restore the prebuilt native binary when install scripts were skipped
Without this the new bunx lane resolves to a command that still fails:
bun skips lifecycle scripts for a bunx fetch, so @ladybugdb/core's
install script never copies lbugjs.node up from its per-platform
sub-package and every native command dead-ends on 'LadybugDB native
binary (lbugjs.node) is missing'.
The existing guidance cannot rescue that case. It offers pnpm
--allow-build, a global install, or adding trustedDependencies to a
project package.json — bunx has no project package.json to add to, no
per-invocation opt-in, and re-extracts the package on every run, so an
out-of-band repair is wiped before the next invocation. In-process
recovery is the only thing that can work.
Recovery is cheap because nothing is actually absent: the binary is
already on disk in @ladybugdb/core-<platform>-<arch>, and the skipped
script only copied it up. Redo that copy (prebuilt only — never a source
build, never a network fetch) before reporting failure. Best-effort by
construction: read-only node_modules, an absent sub-package or an
unsupported platform all fall through to the existing diagnostics
unchanged, which a test pins.
Also covers pnpm dlx without --allow-build and npm --ignore-scripts.
Declare trustedDependencies so a plain `bun install` in this repo
produces a working native binary too — the remedy the error message
already prescribes.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016R9psS9gJ73MRyquoBoPKg
* fix(ai-context): name every install-free runner in the generated bootstrap note
The emitted gitnexus:start block told a reader with no runner yet to run
`npx gitnexus analyze`, falling back to a global npm install. Both name
binaries a bun-only machine does not have, so the generated AGENTS.md and
CLAUDE.md offered it no reachable bootstrap path.
List npx, bunx and pnpm dlx instead of resolving one. The block is
committed, so emitting the command this machine happens to resolve would
make two contributors on different package managers rewrite it at each
other on every analyze — the per-machine churn #1706 removed. Naming all
three keeps the note machine-independent and correct everywhere.
Regenerates this repo's own committed block to match.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016R9psS9gJ73MRyquoBoPKg
* fix(cli): address PR #2765 review — bunx liveness, restore diagnostics, docs
Addresses all five review comments on #2765.
P1 — `hasBun()` was a PATH-existence check only, so a present-but-broken
`bunx` shim (partial uninstall, failed `bun upgrade`) was selected with no
functional validation. Because selecting `bun` also suppresses the npm-11
npx-crash warning, the result was a silent dead end: no diagnostic, and a
`bunx gitnexus@latest analyze` command that only fails at execution time.
Add `probeRuns()` — a real `bunx --version` liveness probe, gated behind the
cheap spawn-free PATH scan so machines with npm/pnpm still pay nothing. It
ignores the output on purpose (a banner or unparseable version still counts
as alive); only a spawn failure, non-zero exit, or timeout rejects. Injectable
via a new `bunRuns` dep so the mode tests stay host-independent.
P2 — the `gitnexus-cli` skill (and both shipped mirrors) still described the
pre-bunx ladder, stranding exactly this PR's audience: a bun-only machine
whose agent bootstraps from that file was told to use npx/npm/pnpm, none of
which exist there. All three copies now name `bunx` in the ladder and the
bootstrap fallback, with a `shipped-skills-sync` fragment assertion so the
gap is CI-caught (these copies are not byte-compared, only the engineering
family is).
P2 — `restorePrebuiltNativeBinary` collapsed every failure into `false`, so an
EACCES/EROFS from `copyFileSync` was indistinguishable from "no prebuilt
sub-package exists". Users on a read-only `node_modules` layer (a baked
container image mounted read-only — a common CI pattern) got the generic
lifecycle-script advice, which cannot fix a non-writable filesystem. Return a
`RestoreOutcome` instead and route `copy-failed` to its own message.
P2 — document that `trustedDependencies` only takes effect for `bun install` /
`pnpm install` run inside this repo: it does nothing for a `bunx` one-shot or
for a consumer's `bun add gitnexus`. The note sits on
`restorePrebuiltNativeBinary` so a future maintainer cannot mistake that
function for redundant and delete the thing the bunx path actually relies on.
P3 — the `binary_missing` bun advice told `bunx` one-shot users to edit a
package.json they do not have, and listed 1 of the 3 packages this package
now trusts. Both repair messages now share one `BUN_REPAIR_LINES` const with
the full package list and a `bun install -g gitnexus` alternative.
Also: shortened the bootstrap note and raised the CLAUDE.md block budget
2900 -> 2950. The note has to name every install-free runner (that is the
point of the bun lane), and main's own growth since this PR's last green CI
had already pushed the generated block over the old ceiling.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015epfxkEMsmHFNVSFQqAkB4
* refactor(cli): simplify the #2765 review fixes
Cleanup pass over the previous commit — no intended behavior change except
the doctor status line noted below.
Reuse: `probeRuns()` duplicated `probeVersion()`'s entire spawn setup — same
argv, timeout, `windowsHide`, and the CVE-2024-27980 Windows-shim workaround —
in a file with two byte-identical committed copies, so the shim rule lived at
four sites. Its docstring's own objection was to the RETURN SHAPE, not to
reuse, so `probeVersion` now returns `{ ran, major, minor }` and `hasBun` reads
`.ran`. Existing callers only read `major`/`minor`, so nothing else changes.
Also dropped a pointless `const runs = () => …` thunk (`&&` already
short-circuits), and deleted a new test that was a character-for-character
duplicate of `falls back to npx when npm is null-absent and pnpm is also
absent` — its cheapest-first-gate rationale moved into that test's comment.
Correctness in the budget comment: the claim that the bun rung is free because
"pnpm is absent there, so its probe never ran" was wrong. `formatAnalyzeCommand`
spawns `pnpm --version` unconditionally when no global `gitnexus` is on PATH —
that spawn IS how pnpm presence is discovered. Real worst case is 5 subprocesses
/ ~8s, and the 8s needs Windows (`shell: true` spawns cmd.exe for an absent
pnpm); on POSIX an absent pnpm ENOENTs in ~1ms. Comment now says that. Likewise
"a machine with npm or pnpm never pays" was wrong for npm 11+ without pnpm —
that IS the rung that pays.
Altitude: `copy-failed` changed only the message text while still returning
`kind: 'binary_missing'`, so `doctor` would have printed "✗ lbugjs.node missing"
directly above a message saying the binary IS present — exactly the
contradiction #2672 removed. Added a `binary_unwritable` kind, a doctor case,
and a `nativeStatusCases` row. The binary-missing message construction moved
out of `checkLbugNative` into `unrestorableBinaryFailure`, typed
`Exclude<RestoreOutcome, 'restored'>` so a new outcome forces a decision
instead of silently inheriting the lifecycle-script advice.
Drift: the trusted-package list was hand-spelled in five places in
native-check.ts, with "matches gitnexus/package.json" asserted only in a
comment. All five now render from one `NATIVE_BUILD_PACKAGES` const (rendered
output is byte-identical), and the test reads the list out of package.json
instead of restating it, so a fourth native package fails the test rather than
silently shipping stale advice.
Finally, replaced the absolute CLAUDE.md block cap with the ratio the two prior
justifications actually appealed to (`< 5465 * 0.55`). Raising 2700 -> 2900 ->
2950 was a ratchet with no ratchet: an absolute cap can only fail on the PR
that adds the character, and the fix is always to nudge the number. Also fixed
a stale runner ladder in skills-steering.test.ts that still omitted bunx.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015epfxkEMsmHFNVSFQqAkB4
---------
Co-authored-by: drdave-flexnteos <revenaugh.david@gmail.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Gergő Magyar <gergomagyar@icloud.com>
Co-authored-by: Gergo Magyar <gergomagyar0@gmail.com>
|
||
|
azizur100389
|
797e4ef8f6
|
fix(ai-context): document CLI graph fallbacks (#2803)
* fix(ai-context): document CLI graph fallbacks Teach generated GitNexus guidance to pair mandatory MCP graph checks with repo-scoped CLI fallbacks so agents can keep working when MCP is unavailable. Co-authored-by: Cursor <cursoragent@cursor.com> * style(ai-context): satisfy Prettier --------- Co-authored-by: Cursor <cursoragent@cursor.com> |
||
|
Gergő Magyar
|
8b5057f325
|
feat(skills): GitNexus Engineering Tool Kits (#2566)
* feat(skills): add ce-plan — GitNexus+PDG implementation-planning skill Adds .claude/skills/ce-plan: a planning-only skill that builds implementation-ready plans from GitNexus graph navigation (query/context/ impact/trace), bounded statement-level PDG slices (pdg_query, impact mode:pdg, explain), and targeted source verification, with a context ledger to prevent repeated reads and a machine-readable implementation context pack (stable contract for a future ce-implement). Whitelisted in .gitignore and registered in AGENTS.md and CLAUDE.md outside the auto-managed gitnexus block. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(skills): apply ce-plan validation findings (tool contract, consistency, conventions) Tool contract: impact mode:'pdg' shape now includes the schema-required direction param; CDG branch sense documented as the result 'label' field (reason is cypher/raw-edge only); explain caveats corrected to its real false-negative classes (cross-function TAINT_PATH is modeled). Consistency: PDG slice homed in working memory (ledger keeps one-liners); depth knob defined and category-overrides-baseline ordering stated; call_depth (consumed by nothing) and content-hash bookkeeping dropped; Never section folded into Hard rules; Phase 3 deduplicated to a pointer; allowed-repeat escalations defined; budget/discard accounting clarified; verification-commands gathering added to Phase 4; open_questions added to the context pack. From scenario runs: plans now pin the verified-at HEAD commit and index freshness in a header, tag claims [verified]/[graph]/[inferred]/[assumed], quote load-bearing tool output, prefer pre-hook-carrying npm scripts, and support an out:<path> destination override; output path defined as the Phase 1 target repo root. Conventions: AGENTS.md 1.9.0 / CLAUDE.md 1.4.0 changelog rows + metadata bumps; future ce-implement qualified as future. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(skills): rename ce-plan → gitnexus-plan; add cross-CLI (Codex) entrypoints Renames the skill dir, frontmatter, output filename convention, plan H1 (GitNexus Engineering Plan), the future executor handle (gitnexus-implement), the .gitignore whitelist entry, and all AGENTS.md/CLAUDE.md references. Follows the pr-swarm-review cross-CLI pattern: SKILL.md is the canonical CLI-neutral spec, AGENTS.md § Engineering planning is the Codex/any-agent entrypoint, and the README documents the optional user-level ~/.codex/prompts/gitnexus-plan.md slash command plus an invocation matrix. Skill prose de-branded from Claude Code (agent-neutral verification layer). Also fixes two post-review README contradictions: the anti-reread claim now names the ledger's allowed escalations, and 'read-only by contract' is now 'planning-only' (the skill writes exactly one repo file — the plan); the scope-creep rule and template §12 now agree on where deferred follow-ups land. Drops the stale plugin-collision limitation. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(skills): document Codex user-level install path for gitnexus-plan Codex discovers SKILL.md skills from ~/.agents/skills (same path the other gitnexus-* skills install to); README now documents the cp install plus the optional ~/.codex/prompts slash-command file, with the prompt body preferring the repo copy and falling back to the user-level install. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(skills): gitnexus-plan freshness gate + active PDG-layer refresh Freshness is now a Phase 1 gate, not advisory: under the default freshness:strict, a stale index is refreshed once per planning session via node .gitnexus/run.cjs analyze --index-only (appending --pdg when the task will reach the PDG phase), then the context resource is re-read. A missing PDG layer likewise triggers the one permitted --index-only --pdg refresh and re-probe instead of a passive recommendation. freshness:accept (or a failed/impractical refresh) preserves the old behavior: plan on the stale graph, source-weighted, labelled in the plan header. --index-only is the load-bearing flag choice — it suppresses all file generation, so the planning-only contract holds (only the .gitnexus store changes). Ledger gains an index_refresh record; plan header states fresh / refreshed / refresh-skipped-with-reason. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(skills): gitnexus-plan runner build check before freshness refresh When the target repo builds the analyzer from its own source (bin → dist/ mapping, as gitnexus/ does), the Phase 1 freshness gate now verifies dist/ is current before running the analyze refresh — rebuilding via the package's build script when any analyzer source file is newer than the built entrypoint — and prefers that freshly built CLI. Otherwise a stale dist re-indexes with outdated extraction logic and the 'fresh' index lies. Rebuilds are recorded in the ledger's index_refresh; the PDG-phase refresh inherits the same check. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(skills): add gitnexus-work executor and gitnexus-lfg pipeline gitnexus-work executes a gitnexus-plan as verified atomic commits: consumes the §11 implementation_context pack, drift-checks the plan's evidence pin against HEAD, re-verifies assumptions before relying on them, runs impact before every symbol edit and detect_changes before every commit (repo mandates), builds tests from the plan's scenarios, and routes structural drift back to gitnexus-plan Deepen mode instead of coding around it. gitnexus-lfg is a thin orchestrator: gitnexus-plan → blocking user gate (deepen / proceed / stop, deepen loops allowed) → gitnexus-work → review via the existing gitnexus-pr-review skill (open PR, else branch diff vs default). One bounded fix cycle for review findings; never pushes or opens a PR on its own. gitnexus-plan gains a Deepen mode (re-run freshness gate, escalate to depth:deep, re-verify graph/inferred/assumed claims toward verified, rewrite the same file); its 'future gitnexus-implement' placeholder is retired in favor of gitnexus-work. Registered via .gitignore whitelists, AGENTS.md 1.10.0 (section renamed to Engineering planning & execution), CLAUDE.md 1.5.0. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(skills): apply cross-skill review findings to the gitnexus skill family Two P1s: gitnexus-plan Deepen mode now re-anchors before re-pinning (diffs the old evidence pin over every [verified]-claim file and re-reads or downgrades before the header moves — moving the pin without this laundered stale claims as verified); the index-refresh budget is stated once in Phase 1 (one --index-only refresh plus at most one Phase 3 --pdg upgrade per session, Deepen = its own session) with ledger and pdg-slice deferring to it. Contract fixes: gitnexus-work's drift check now covers every file the pack cites (not just files_to_modify) and parses the full pack incl. primary/related symbols and acceptance_criteria (walked in Phase 4 alongside §13); a pre-completed check skips §7 steps already landed and Deepen gains a reconcile-execution-state step, closing the mid-execution route-back loop; pack assumptions must name what to check and how. lfg: Lane 4 passes the merge-base to detect_changes compare (two-dot diff misattributes upstream commits when default advanced), branch-diff is the stated normal case, oversized review findings route to the plan gate instead of overflowing direct mode, the one-fix-cycle cap is explicit on re-run, and headless runs end at the plan gate with the plan as deliverable. work: blank mode narrowed to *gitnexus-plan*.md with a re-execution guard, direct-mode discipline spelled out, branch meaningfulness defined against the plan slug, and the plan document is committed as the branch's docs commit (review diff includes it). Planning-only contract now names the dist/ rebuild as the second permitted state change; Phase 5.1 names the four claim tags; stale AGENTS.md anchors fixed. Known latent issue left untouched: gitnexus/gitnexus-pr-review pairs a three-dot example with a two-dot detect_changes compare — that skill is also shipped by the plugin, so fixing it here would drift the copies; lfg compensates by passing the merge-base. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(skills): ship the engineering skill family with the gitnexus package npm i -g gitnexus users now get gitnexus-plan / gitnexus-work / gitnexus-lfg: the three skills are added to gitnexus/skills/ in directory form (SKILL.md + references/), which installSkillsTo already enumerates dynamically and copies recursively to every editor target (~/.agents/skills for Codex, Cursor, OpenCode, Qoder, ...) on gitnexus setup — uninstall enumerates the same root, so removal stays clean. The Claude Code plugin channel (gitnexus-claude-plugin/skills/) carries the same copies plus the standard per-skill mcp.json. Global-install support in the skill text: gitnexus-plan Phase 1 now resolves the analyzer runner explicitly — node .gitnexus/run.cjs analyze when the project has a runner, else gitnexus analyze (installed CLI), else npx gitnexus analyze — and all analyze mentions route through it, satisfying the skills-steering policy (#1939/#1945) which sweeps the plugin copies. New drift guard test/unit/shipped-skills-sync.test.ts asserts the npm and plugin copies stay byte-identical to the canonical .claude/skills/ family (plugin = canonical + mcp.json), same discipline as run.cjs ↔ resolve-invocation.ts. skills-steering + shipped-skills-sync: 11/11 green. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(eval): workflow_bench — measure the skill workflow's token savings Benchmarks gitnexus-plan → gitnexus-work against a baseline agent (--disallowedTools Skill) on identical tasks, in fresh detached worktrees, using real headless Claude Code sessions; every number comes from the CLI's --output-format json usage report (field names validated against a live 2.1.207 session). Reports per-arm medians (input/cache/output tokens, cost, wall time, turns), a savings row, and resolve status from a per-task verify command — savings on failed tasks are flagged, not celebrated. Per-task setup hook prepares fresh worktrees (deps); --permission-mode bypassPermissions (default) lets sessions run unattended in the throwaway trees. Free-model support: --base-url/--auth-token/--model route headless sessions through any Anthropic-compatible endpoint; free-model.litellm.yaml is a ready litellm-proxy template for OpenRouter :free variants or local Ollama, so benchmarking burns no paid tokens (README documents rate limits and the small-model skill-following caveat). Harness validated end-to-end with a stub CLI (worktree lifecycle, both arms, plan→work chaining, verify, aggregation, report) and 4 pytest units for the pure aggregation/savings/report helpers. AGENTS.md 1.11.0. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(eval): record first workflow_bench calibration run Trivial-task calibration (add -V alias): both arms resolved; workflow arm ~4.3x baseline cost — the documented overhead-dominated regime, recorded so the regime boundary is empirical rather than asserted. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(eval): workflow_bench scenario matrix — arm variants, task classes, churn Ground-base measurement across scenarios: tasks.scenarios.yaml spans four labeled classes (trivial → investigation-bug → investigation-feature → cross-module) with deterministic verifies (prescribed test files). New arms: workflow_direct (gitnexus-work direct mode — the middle option that locates the routing boundary lfg's gate and work's triage encode) and baseline_nomcp (no skills AND no graph tools — separates workflow-discipline value from GitNexus-tool value; off by default). Records now carry task class and diff churn (files/+ins/−del vs the starting commit) as an over-engineering proxy; the report renders a class column and per-arm savings rows vs baseline. 5 pytest units + stub-CLI e2e of the full three-arm matrix. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(eval): record workflow_bench ground base; fix churn measurement bias Ground base (3 classes x 3 arms, n=1/cell): every arm resolved every task — pass/fail quality saturates at this difficulty, making the comparison pure cost. Full plan→work never amortized its ~$9-11 fixed cost on tasks a baseline finishes in ≤35 turns (−211% to −333% cost); workflow_direct sits near baseline (−15% to −55%, once faster wall) with more test coverage. Routing implication recorded: direct mode/plain agent below this scale, full workflow for cross-module / multi-session / plan-as-deliverable work. The cross-module cell and multi-run variance are the next measurements. Churn fix: git add --intent-to-add -A before diffing (arms that never commit no longer undercount new files) and :(exclude)docs/plans (the committed plan doc no longer inflates workflow churn); this run's churn numbers predate the fix and are omitted from the recorded table. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * perf(skills): cost-optimize the workflow from measured ground base Every optimization targets a measured fixed-cost component (eval/workflow_bench ground base: workflow arm −211% to −333% vs baseline, all tasks resolved): - Plan form is category-priced: compact form (core sections w/ § anchors preserved, ≤80 lines excl. pack, mini-pack subset of the context pack) for narrow/default categories; the full 13 sections only for deep work (refactor/security/performance/concurrency/architecture). A compact plan outgrowing its cap reclassifies to full rather than overflowing. - Freshness gate is category-priced: compact categories default to accept (source-weighted, refresh only when a graph claim becomes load-bearing); strict stays the default for full-plan categories — the rebuild+re-index was the largest single fixed cost. - Turn economy: per-category tool-call budgets (~10 to ~45; architecture uncapped); budget exhaustion routes open questions to §12 instead of more digging. - gitnexus-work fast path: HEAD == evidence pin → skip all citation re-reading (the pin's entire point); mini-pack fields tolerated. - lfg Lane 1 boundary triage: tasks below the measured ~35-turn boundary get offered gitnexus-work direct mode before the plan lane is spent. Copies re-synced (npm skills/, plugin, ~/.agents); steering + sync guards green. Re-measurement of the workflow arm follows to verify the numbers actually improve. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(eval): record optimization re-measurement — inv-bug workflow cell −20% cost Same task, same conditions, post-830a0459 skills: $14.56→$11.70 (−20%), 83→72 turns, cache_read −24%; verified in-transcript that the compact form, turn budget, and skipped rebuild/re-index all fired. Wall +15% from a work- session test-debugging tail (n=1 variance). Regime unchanged (~3.5x baseline on this class) — routing rule stands. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(eval): per-arm clone isolation — worktree ref-namespace leak contaminated an arm The cross-module workflow_direct cell reported an impossible 28-turn solve with churn byte-identical to the workflow arm: git worktree add shares the repo's ref namespace, so the workflow arm's slug branch (created by gitnexus-work Phase 2) survived worktree removal and the direct arm found and adopted the completed work. Arms now get isolated git clone --shared copies (object store via alternates, refs clone-local — agent branches and stashes die with the clone; origin/<ref> fallback for non-default refs). Leaked branch deleted; baseline arm verified clean (0 branch references in its transcript); cell marked invalidated pending re-run. Records the valid cross-module cells: workflow $18.32 vs baseline $18.03 (premium −1.6%, vs −211%..−333% on smaller classes) — fixed costs amortize at this scale, with a less destructive diff and a plan artifact as bonus; resolve rate still tied. Churn fingerprinting is what caught the contamination — noted in the README as an integrity check. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(eval): complete cross-module cell — direct mode wins 47% cost / 56% wall Clean clone-isolated re-run: workflow_direct resolved the hardest class at $9.53/52 turns/15m vs $18.03/98/34m baseline and $18.32/107/37m full workflow. The measured story across all four classes: the execution discipline (gitnexus-work) is the consistent sweet spot and delivers real token savings on hard tasks; the planning pass buys its artifact, not same-session savings. Resolve rate tied everywhere (n=1/cell caveat). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(eval): add trajectory-gated skill evolution (#2431) - Pair prompt candidates with incumbent workflow arms - Gate promotions on pinned-model quality and efficiency - Expire router evidence and document its lifecycle * fix(eval): allow pr-review skill candidates * feat(skills): rename and generalize GitNexus review * feat(eval): external-comparator and review arms for workflow_bench - ce_workflow / ce_workflow_direct: compound-engineering ce-plan/ce-work arms prompted with the same structure as the gitnexus arms - review / ce_review: gitnexus-review vs ce-code-review on an identical diff applied by the task's setup - plan handoff is snapshot-based: committed example plans in docs/plans/ tie on clone mtimes and broke the name-glob pick (executed a stale plan) - verify output tail is recorded per run and the final working-tree patch is kept, so failed rows are diagnosable after the clone is destroyed Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(skills,eval): address #2431 review — data-safe rename migration, fail-closed bench evidence - setup: never delete a legacy renamed skill dir — the installer cannot prove ownership (users customize or hand-write skills under these names); warn with the path instead, and the test now asserts survival - workflow_bench: fail closed when a session's --output-format json report is empty, malformed, or missing usage fields — an exit-0 shell with no parseable usage no longer counts as measured evidence (5 parametrized regression tests) - workflow_bench: document the trust model prominently (task setup/verify are shell-executed, sessions run bypassPermissions with the parent env, candidate overlays are prompt injection surface) in README + docstring - free-model.litellm.yaml: master_key from LITELLM_MASTER_KEY env instead of a static token; loopback-binding warning - ci: run the eval workflow_bench pytest suite on ubuntu (pytest+pyyaml only — no full eval stack) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(eval): demand observed foreground verification in headless work-arm prompts In a headless -p session there is no later turn: a work arm backgrounded its slow test run, scheduled wakeups that can never fire, and reported done while two of its tests failed. All four work-arm prompts (both skill families, symmetric) now require verification output to be observed inside the session. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(skills): ask plan depth up front instead of offering deepen afterwards gitnexus-plan Phase 0 now asks one blocking question in interactive sessions — quick / standard / deep, mapped onto the existing depth/form/ freshness knobs — when the invocation carries no explicit depth signal. Explicit knobs and headless runs skip the question (category posture unchanged, so benchmarks and automation behave as before). gitnexus-lfg's plan gate slims to proceed/stop: depth was already the user's up-front choice, so deepening is no longer offered by default — an explicit deepen request at the gate and executor route-backs still run Deepen mode, which remains the mechanism for strengthening an existing plan document. All shipped copies resynced (npm skills/, Claude plugin); AGENTS.md 1.13.0 and CLAUDE.md 1.7.0 pointers updated, including the analyzer's regenerated index-stats block at this branch's head. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(skills): taint pass, expert lenses, and post-work index refresh gitnexus-review gains a PDG-backed taint-and-dependence pass (explain + pdg_query, --pdg folded into the stale refresh on trust-boundary diffs) and an Expert lenses section: domain reviewers derived from the graph's clusters plus four cross-cutting lenses (architectural fit, language conformance per the repo's own contract, Definition of Done, simplicity), dispatched once after the evidence-gathering steps and scaled to the diff. gitnexus-work Phase 4 now refreshes the knowledge graph after the DoD walk via the resolved-runner ladder with analyze --index-only, so the lfg review lane and later sessions query the finished work without dirtying the tree. lfg's threshold-governance paragraph moves to its README; eval citations are tagged as measured in the GitNexus repo. All shipped copies re-synced. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(cli): remove legacy gitnexus-pr-review on uninstall; cover the rename migration uninstall's removal set now includes LEGACY_SKILL_DIR_NAMES derived from RENAMED_SKILL_DIRS, so a pre-rename install is cleaned up instead of orphaned. The rename warning gains behavioral coverage (fires with a legacy dir present, silent without), and shipped-skills-sync asserts legacy names stay absent from every shipped tree. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(eval): metric provenance, error-kind rows, skill-invocation verification, gate noise floor The promotion gate defaults to cost_usd (the only metric that includes subagent spend); token metrics carry an explicit main-loop-only warning in the report and promotion.json. Rows are classified by error_kind (session-error / verify-failed / infra-error), excluded from efficiency medians, and the gate requires equal valid-run counts. Each session's transcript is scanned for the expected Skill invocation and fails closed on a verified miss; a one-run resolution edge no longer promotes (noise floor). Per-run timeouts and setup failures record an infra-error row instead of aborting the sweep. Overlays touching skills no candidate arm exercises are rejected up front. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs: fix skill routing paths, version headers, and skill rosters Routing tables point at the tracked direct skill paths (matching the post-#2434 generator output), AGENTS.md/CLAUDE.md headers match their latest changelog rows, the 1.12.0 row describes what the migration actually does, package/cursor READMEs list the full shipped skill roster, and the swarm READMEs describe /gitnexus-review's expert lenses instead of calling it single-agent. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * ci: drift-guard workflow for skill copies; pin eval pip deps; track docs/plans ci.yml ignores '**.md', so an md-only skill edit would merge without the shipped-skills-sync test running — skill-sync.yml triggers exactly on the guarded trees. The eval job's pip install is version-pinned, and docs/plans/ is unignored so gitnexus-plan output can be committed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(ci): keep the runner-invocation literal in gitnexus-review; add concurrency block to skill-sync skills-steering requires skills with a stale-index hint to carry the exact 'node .gitnexus/run.cjs analyze' form — restore it with the fallback ladder as a parenthetical instead of replacing it. skill-sync.yml gains the top-level concurrency block the workflow-convention check enforces. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(skills): token-economy guidance for expert lenses Merge lenses that ground in the same material into one reviewer, and use cheaper model/effort tiers for mechanical lenses where the harness offers them, reserving the strongest engine for adversarial judgment. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test(eval): isolate transcript home on Windows Ensure workflow_bench transcript tests set USERPROFILE alongside HOME so Path.home() resolves to the temporary test home on Windows. * docs(skills): fold PR #2522 execution learnings into review/work/plan Eight incident-backed hardenings from running the full skill cycle (review -> plan -> work, 28-finding fix series) on PR #2522: gitnexus-review: - Expert lenses execute the code under review on candidate failing shapes (empirical probe outranks source reading — every HIGH the language lenses found came from a probe, not a read). - Step 7 re-runs the exact CI check for refreshed baselines/fingerprints (a stale committed artifact is invisible in the diff; caught a red benchmarks arm). - Step 8 treats version/invalidation constants as review surface (INCREMENTAL_SCHEMA_VERSION class recurred verbatim from #2494). gitnexus-work: - Step 4 proves regression tests discriminate against the pre-fix tree. - Step 5 rebuilds executed build output before every verification run (parse workers load dist/; a correct fix 'failed' until rebuilt). - Step 6 makes stage -> detect_changes -> commit one unbroken sequence. gitnexus-plan: - Phase 0 seeded-evidence mode: plan FROM a completed review's verified findings instead of re-running the graph ladder. - Template §7: fingerprint/golden-guarded output rebaselines once, at the series tip. All distribution copies resynced; shipped-skills-sync + skills-steering 24/24 locally. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(eval): close the skill-evolution loop with an automated proposer driver workflow_bench.evolve adds the three arrows the README described as manual: a proposer session that turns loser trajectories (results.jsonl rows, transcripts, patches, the learning queue) into ONE bounded candidate overlay, a driver that iterates propose -> paired benchmark -> deterministic gate up to --generations, and an --apply step that copies a promoted overlay onto the canonical skills and shipped mirrors as a working-tree diff. The trust boundary is unchanged: overlays re-validate through candidate_overlay_files before any benchmark or apply consumes them, and committing, CI, and the PR merge stay human. learnings.jsonl is gitignored: it is machine-local evidence, like the session transcripts it complements. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(skills): route live-task friction into the evolution learning queue Each family skill gains a short 'Skill feedback' section: on friction with the skill's own instructions, append one JSON line to eval/workflow_bench/learnings.jsonl (GitNexus repo only) — never self-edit the skill from a live task. The proposer in workflow_bench.evolve consumes the queue as hints; a learning reaches a shipped skill only by beating the incumbent on the paired benchmark. All shipped mirrors re-copied byte- identical. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * ci(tests): run the evolve helper tests in the eval pytest job test_evolve.py needs only pytest+pyyaml, same as the harness tests the job already runs — without this line the new module had no CI coverage. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(ci): comment-triggered GitNexus review agent for PRs '@gitnexus review' from a maintainer (OWNER/MEMBER/COLLABORATOR; the action re-validates write access) runs the repo's gitnexus-review skill headlessly against the PR and posts the review as a sticky comment — remote triggering with no local setup. Read-only by construction: contents: read token, Write/Edit and web tools disallowed, Bash allowlisted to git reads and the gitnexus CLI; analyze parses PR code with tree-sitter, never executes it. Requires the ANTHROPIC_API_KEY repository secret; activates once the file is on the default branch. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(ci): dispatch lane + existing OAuth secret for the review agent Align with claude.yml: same action pin and the CLAUDE_CODE_OAUTH_TOKEN secret the repo already carries — no new secret to configure. Add a workflow_dispatch lane (PR number input) so the agent can be triggered from the Actions UI and tested before the issue_comment trigger reaches the default branch. Allowlist gh pr view/diff and gh api, which the review skill uses to pin PR SHAs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix(ci): close a fork-PR RCE vector in the review agent's tool allowlist A live headless run of the exact workflow session against PR #2431 (66 turns, full gitnexus-review pass) surfaced a real HIGH-severity confused deputy: .gitnexus/ is gitignored, not blocked — a fork PR can commit its own .gitnexus/run.cjs, issue_comment checks out PR-head content, and the skill's runner ladder tries 'node .gitnexus/run.cjs analyze' first. That would execute fork-controlled JS inside a job holding CLAUDE_CODE_OAUTH_TOKEN and a write-scoped GITHUB_TOKEN — the opposite of the 'PR code is read, never executed' claim in the workflow's own header. Fix: drop the run.cjs allowlist entry so analyze always resolves through npx gitnexus (npm registry, not the checked-out tree); the skill's documented fallback mode covers the resulting graceful degradation. Also drop 'gh api' (not read-only — accepts -X POST/PATCH/DELETE) and downgrade pull-requests: write to read (comment posting only needs issues: write; the prompt already forbids formal review submission). Same session flagged a latent evolve.py bug: select_evidence's cost sort used dict.get's missing-key default, which doesn't cover an explicit JSON null in a foreign --seed-results row and crashes proposer setup with TypeError. Guarded with 'or 0.0' and added a regression test. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * fix: harden PR review and evolution trust boundaries * ci: follow workflow concurrency convention * fix(eval): make terminating error paths explicit * fix: unblock hardened review runtime checks * test: make containment canaries deterministic * test: expose Claude canary tool failures * fix: adapt clean shell environment for Claude * fix(eval): accept the runner's transcript source key in evidence preflight The proposer evidence preflight required transcript-artifact metadata to be exactly {path, sha256, bytes}, but the runner stamps a fourth provenance key (source=parent-captured-stream-json). Any --seed-results or generation>=2 run therefore aborted with SandboxError before proposing or promoting. Pin the producer literal as PARENT_EVENT_STREAM_SOURCE and validate it in the metadata check, and round-trip real producer output through sum_sessions into the preflight so the schema can't drift again. * fix(eval): treat an unmeasured session cost as unavailable, not $0 well_formed validated only the nested usage block, so an otherwise-successful session missing total_cost_usd was recorded as cost_usd=0.0 — and cost_usd is the default promotion metric (lower wins), so a cost-less session scored as free and could win promotion it never earned. Extract cost via measured_cost() (None on absent/garbage, a measured 0.0 preserved), propagate None through sum_sessions/aggregate/savings/report, and have the gate refuse to rank on a metric that was not measured on every run in both arms. * fix(eval): warn when ranking on the main-loop-only num_turns metric num_turns comes from the CLI's top-level usage (main-loop session only), like output_tokens, but selecting it emitted no metric_warning — so a subagent-heavy candidate could look artificially efficient. Add num_turns to MAIN_LOOP_ONLY_METRICS and broaden the warning to cover turns. * fix(eval): fail closed when an overlay adds a file with no committed base An overlay adding a new .md under gitnexus-{plan,work} passes the structural overlay checks but has no committed base for committed_destination_base_digests to bind against, so it raised an uncaught ValueError that crashed the evolve driver (and runner --candidate-overlay) mid-run. Catch it at both call sites: evolve reports NOT PROMOTED and exits, runner routes it through parser.error. * feat(eval): circuit-break the runner sweep on a systemic outage A sustained upstream outage used to pay out every remaining --timeout window one session at a time. Track consecutive session/infra/cleanup failures via a pure systemic_outage_streak helper; after --outage-streak (default 5) in a row, stop the sweep, still write report.md/promotion.json from partial evidence, and exit non-zero so evolve.py halts instead of proposing from truncated evidence. A task's own resolved=False never trips the breaker. * fix(cli): report a dirty working tree as stale in gitnexus status status --json (and the human output) computed up-to-date from commit + runner identity + completeness only, so a repo with uncommitted source changes at a matching HEAD was reported up-to-date while analyze would still re-index it. A graph-backed agent gating on that JSON could skip re-analysis on a stale graph. Extract analyze's dirty-tree check into a shared isWorkingTreeDirty() in storage/git and fold it into the status freshness decision. * fix(ci): use single-slash deny globs in the review agent's disallowedTools github.workspace already expands to an absolute path, so Read(/${{ github.workspace }}/**) and Read(//proc/**),(//sys/**),(//dev/**) produced double-slash patterns that a normalizing matcher may not match — silently no-opping the deny layer. Not exploitable (the allowlist is the primary control and never grants those paths), but the globs should be well-formed. Update the pinned test strings. * ci: install gitnexus-shared with npm ci from the committed lockfile The gitnexus-shared build floated its deps via npm install in three workflows (skill-sync, ci-tests, and — most importantly — the release publish.yml) while every other install step uses npm ci. The lockfile is committed and in sync, so switch all three to npm ci for reproducible, locked installs. * test(cli): make the shipped-skills drift guard reject symlinks listFilesRecursive walked with readdirSync and snapshotDir read with readFileSync, both of which follow symlinks — so a mirror file symlinked to the canonical tree passed the byte-compare (and a symlinked mirror dir would be followed too). Reject a symlinked root via lstat and any symlinked entry via Dirent.isSymbolicLink, with negative tests (skipped on Windows). * test(eval): guard the candidate-skill vs mirror-root coverage invariant MIRROR_SKILL_ROOTS omits the Cursor tree, safe only because no candidate skill is cursor-shipped. Pin that invariant: every CANDIDATE_SKILLS entry must exist under canonical + every mirror root and must not ship to Cursor, so adding a cursor-shipped skill to the candidate set (the PR #2488 asymmetric-sync class) fails loudly instead of syncing three of four trees. * docs(ci): describe the review agent's staged post-merge rollout The DoD asked for a dry-run or triggered run before merge, but an issue_comment (or newly added workflow_dispatch) workflow only ever executes the default-branch copy, so it cannot be exercised from the PR that introduces it. Reword the DoD and the activation checklist to a staged rollout: merge registered-but-disabled, validate same-repo and fork execution post-merge, then enable the variable. * fix: pin plugin skill mcp.json to the release version via #2445 tooling The ten plugin skill mcp.json launched `npx -y gitnexus@latest mcp` on every skill connect — non-reproducible and a supply-chain surface, and (unlike the persisted setup config) never pinned. Extend sync-plugin-manifests.mjs with an mcp surface kind that stamps the gitnexus@<version> launch arg, pin all ten to 1.6.9 now, and keep them byte-identical so the drift guard stays green. The release lifecycle + publish.yml --check now re-stamp them like the four manifest surfaces; only READMEs stay on @latest as docs. * test(eval): prove the proposer's built-in file tools are confined The real-Claude canary only exercised Bash + MCP, so it proved process/MCP containment but not that the proposer's built-in file tools stay inside their mounts. Add a canary over the exact PROPOSER_ALLOWED_TOOLS surface and the same read-only /evidence mount as run_proposer (allowlist extracted to a shared constant so it can't drift): Read reaches /evidence, a Write into the read-only evidence mount is denied, and a Write lands in the output tree. * fix(eval): apply the candidate overlay after task setup for fair arms The candidate overlay was applied before the task's untrusted setup ran, so setup could observe candidate prose and the incumbent/candidate arms started from different pre-overlay state. Reorder within the sandbox: capture the base (pre-overlay) skill digest, run setup against the base skills, verify setup did not tamper them, then apply the overlay and capture the post-overlay digest the model must preserve. apply_candidate_overlay stages path-specific overlay files, so setup's uncommitted changes stay out of the baseline and churn is unchanged. Graph freshness for the review arm is handled by the status dirty-tree fix plus the review skill's stale-triggered re-index, not by reordering the cached per-task-sha graph materialization (which is mechanically blocked). * test(eval): end-to-end containment proof of the autonomous proposer Drives the real run_proposer through bubblewrap with a deterministic scripted model (no paid API): it reads the read-only evidence bundle and writes a candidate gitnexus-plan skill edit plus a rationale into the sandbox output tree; run_proposer enforces the trust boundary and copies only the validated overlay + proposal out. This exercises the autonomous-proposal stage of the self-evolution loop end-to-end in the eval/containment CI job (the gate and apply stages are covered by test_workflow_bench_evolution and test_promotion_apply). Env-gated on GITNEXUS_REQUIRE_CLAUDE_CANARY, so it runs only where the pinned Claude binary and user namespaces are available. * fix(eval): let the proposer author its overlay via Bash Running the end-to-end proposer canary in the containment CI job surfaced a real bug: run_proposer starts the session with --bare, which hard-disables the Write/Edit tools ("Write exists but is not enabled in this context"), yet allowlisted Edit/Write and omitted Bash. The proposer therefore had no working way to write its candidate overlay — the self-evolution loop could never produce a candidate. The sandbox settings already pre-authorize Bash (autoAllowBashIfSandboxed) and confine writes to workspace/tmp/home, so switch PROPOSER_ALLOWED_TOOLS to Read/Grep/Glob/Bash and tell the proposer to author files with Bash. The end-to-end test now drives the real run_proposer through bubblewrap and asserts a validated overlay + proposal are produced (this also replaces the earlier file-tool canary, whose Write/Edit premise was moot). * test(eval): author the proposer overlay with newline-free Bash content The nested shell-sandbox prefix mangles embedded newlines, so the multi-line overlay content never landed. Use single-line content for the deterministic proposer canary. * test(eval): drop the unverifiable end-to-end proposer canary The scripted proposer overlay never materialized in the containment job across runs, and the model tool-result content is not visible in CI logs, so the test cannot be finalized without an environment where the sandbox can actually run. Keep the verified production fix (Bash-authoring in run_proposer); the proposer sandbox/containment stays covered by the existing Bash+MCP and process-tree canaries. * test(cli): drop run-analyze.ts from the windowsHide spawn-family list U7 moved run-analyze.ts's only child_process call (the git status --porcelain dirty check) into storage/git.ts (already covered by this test, with windowsHide). run-analyze.ts no longer imports a spawn-family function, so the windowsHide-regression test's 'must have >=1 spawn call' invariant failed for it. Remove it from SRC_FILES. --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> Co-authored-by: Zander Raycraft <zanderjraycraft@gmail.com> Co-authored-by: Azizur Rahman <azizur100389@gmail.com> |