diff --git a/run.json b/run.json
index d8247652e..bff631990 100644
--- a/run.json
+++ b/run.json
@@ -504,7 +504,7 @@
     "kind": "running"
   },
   "status_updated_at": "2026-07-01T16:19:34.284021538Z",
-  "last_event_at": "2026-07-01T16:22:07.464687848Z",
+  "last_event_at": "2026-07-01T16:44:45.127309562Z",
   "pending_control": null,
   "checkpoints": [
     {
@@ -689,9 +689,9 @@
       }
     },
     {
-      "seq": 0,
+      "seq": 49,
       "checkpoint": {
-        "timestamp": "2026-07-01T16:24:40.579584914Z",
+        "timestamp": "2026-07-01T16:24:43.673691189Z",
         "current_node": "preflight_lint",
         "completed_nodes": [
           "start",
@@ -700,26 +700,131 @@
           "preflight_lint"
         ],
         "node_retries": {},
+        "context_values": {
+          "internal.retry_count.preflight_lint": 0,
+          "command.output": "blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126",
+          "current_node": "preflight_lint",
+          "internal.work_dir": "/home/daytona/workspace/fabro",
+          "failure_class": "",
+          "graph.rankdir": "LR",
+          "internal.fidelity": "compact",
+          "graph.model_stylesheet": "\n            * { model: claude-opus-4-8; }\n        ",
+          "internal.node_visit_count": 1,
+          "outcome": "succeeded",
+          "thread.preflight_compile.current_node": "preflight_lint",
+          "thread.start.current_node": "toolchain",
+          "internal.retry_count.start": 0,
+          "internal.run_id": "01KWF7MM3VPXZZA8BTHJXE9VT1",
+          "internal.retry_count.preflight_compile": 0,
+          "graph.goal": "# Plan A — `SecretRedactor` in `fabro-redact`\n\n**This is Plan A of three** (split for parallel execution):\n\n- **Plan A (this file)** — add a per-run secret-value redactor to `fabro-redact`.\n  Self-contained; touches only `fabro-redact/`. **Run in parallel with Plan B.**\n- **Plan B** — resolve `secrets.*` tokens at the run boundary. **Run in parallel\n  with Plan A.**\n- **Plan C** — wire redaction across leak surfaces + hooks. **Run after A and B\n  merge** (it consumes this crate's type and Plan B's lookup).\n\nThis plan is inert on its own: it adds a tested library primitive that Plan C\nwires up. Shipping it alone changes no behavior.\n\n> **Token notation.** Interpolation tokens are written in this file without their\n> enclosing double curly braces, so the file is safe to pass directly as a\n> workflow goal (the goal templater would otherwise try to expand them). Read\n> `secrets.NAME`, `env.NAME`, and `secrets.*` as the double-curly-brace-wrapped\n> token form used everywhere else in the codebase, and write the real\n> double-brace syntax in the code, tests, and docs you produce.\n\n---\n\n## Overall goal (shared context)\n\nMake secret tokens (`secrets.NAME`) in workflow config resolve from the server\nvault, at the run boundary, with values that never get persisted, never leak, and\nfail closed when a secret is missing or the wrong type. The redaction guarantee\nfor declared secrets is: content-based redaction (already present) is the\nuniversal baseline, plus a per-run registry of resolved secret **values** so a\ndeclared secret is redacted even when it does not look like a credential. **This\nplan builds that registry primitive.**\n\nWhy per-run and not a process global: a test-only in-process run path executes\nmultiple runs in the same process, so redaction state must be per-run, never a\n`static`/global.\n\n## Conventions\n\n- **TDD.** Write the failing test first, then the code.\n- Match the codebase: Rust import style (types by name, functions via parent\n  module, no glob imports in production), `strum` for enum string maps, keep\n  test-only helpers behind `#[cfg(test)]`.\n- Plain-English commit messages, PR text, and comments — no internal planning\n  identifiers.\n- The verify gate runs nightly `fmt --check`, nightly\n  `clippy --all-targets -D warnings`, `cargo nextest run --workspace`, docs check,\n  web/api-client typecheck, and a release build. Implement so all pass.\n- Never print or log a secret value.\n\n---\n\n## Implementation\n\n### A.1 — Add the `SecretRedactor` type\n\nFile: new `lib/crates/fabro-redact/src/secret_registry.rs`, exported from\n`lib/crates/fabro-redact/src/lib.rs`.\n\nAdd a cheap, cloneable, per-run registry of secret values that redacts exact\nmatches regardless of shape. It composes *after* the existing content-based\nredaction (`redact_string`, `redact_json_value`) — this type does not replace\nthem.\n\nShape:\n\n- `SecretRedactor` backed by shared, interior-mutable state (e.g.\n  `Arc<Mutex<Vec<String>>>` or `Arc<RwLock<...>>`) so a clone handed to a\n  different subsystem observes registrations. Derive `Clone` and `Default`; an\n  empty redactor is a pure no-op.\n- `fn register(&self, value: impl Into<String>)` — store a secret value to be\n  redacted. **Ignore empty or whitespace-only values** (registering an empty\n  string would turn all output into `REDACTED`). De-duplicate.\n- `fn redact_into(&self, s: &str) -> String` — replace every registered value\n  substring with the same `\"REDACTED\"` marker used by `redact_string`. Replace\n  **longest values first** so a secret that is a substring of another is handled\n  correctly. If the registry is empty, return the input unchanged (fast path).\n- `fn redact_json(&self, value: serde_json::Value) -> serde_json::Value` — walk\n  the JSON tree and apply `redact_into` to every string leaf (both object values\n  and array elements; object keys are left as-is). Exact-value matching is\n  unambiguous, so unlike `redact_json_value` this pass does not skip any keys.\n- Optional `fn is_empty(&self) -> bool` for callers that want to skip work.\n\nReuse the crate's existing `\"REDACTED\"` replacement marker (see `redact_string`\nin `lib.rs`) rather than introducing a new literal.\n\n### A.2 — Tests (unit, in the new module)\n\n- A **low-entropy** value (e.g. `\"staging\"`) that `redact_string` would *not*\n  catch is replaced with `REDACTED` by `redact_into` after `register(\"staging\")`.\n- Registering `\"\"` or `\"   \"` is a no-op: `redact_into` leaves unrelated text\n  intact (guard against the empty-value footgun).\n- Overlapping values: register both `\"abc\"` and `\"abcdef\"`; `redact_into` on a\n  string containing `\"abcdef\"` redacts the whole token (longest-first), not just\n  the `\"abc\"` prefix.\n- Empty registry: `redact_into` and `redact_json` are the identity.\n- `redact_json` redacts a registered value nested inside an object value and\n  inside an array element.\n- A clone of the redactor observes values registered through the original (shared\n  state), proving it can be handed to another subsystem.\n\n### A.3 — Verify\n\n- `cargo +nightly-2026-04-14 fmt --check --all`\n- `cargo +nightly-2026-04-14 clippy --workspace --all-targets -- -D warnings`\n- `cargo nextest run -p fabro-redact`\n- release build (`cargo dev build -- -p fabro-cli --release`)\n\n## Dependencies\n\nNone. Parallel-safe with Plan B. This type is consumed by Plan C.\n",
+          "internal.retry_count.toolchain": 0,
+          "internal.thread_id": "preflight_compile",
+          "thread.toolchain.current_node": "preflight_compile",
+          "failure_signature": ""
+        },
+        "node_outcomes": {
+          "preflight_compile": {
+            "status": "succeeded",
+            "context_updates": {
+              "command.output": "blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126"
+            },
+            "notes": "Script completed: cargo check -q --workspace 2>&1",
+            "usage": null,
+            "timing": {
+              "wall_time_ms": 0,
+              "inference_time_ms": 0,
+              "tool_time_ms": 143940,
+              "active_time_ms": 143940
+            }
+          },
+          "preflight_lint": {
+            "status": "succeeded",
+            "context_updates": {
+              "command.output": "blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126"
+            },
+            "notes": "Script completed: cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1",
+            "usage": null,
+            "timing": {
+              "wall_time_ms": 0,
+              "inference_time_ms": 0,
+              "tool_time_ms": 153109,
+              "active_time_ms": 153109
+            }
+          },
+          "start": {
+            "status": "succeeded",
+            "usage": null
+          },
+          "toolchain": {
+            "status": "succeeded",
+            "context_updates": {
+              "command.output": "blob://sha256/fc14b2ba2d770e5cd3169df7a29525c962adfc4cfa3097b9098c63ebd61a748c"
+            },
+            "notes": "Script completed: command -v cargo >/dev/null || { curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y && sudo ln -sf $HOME/.cargo/bin/* /usr/local/bin/; }; cargo --version 2>&1",
+            "usage": null,
+            "timing": {
+              "wall_time_ms": 0,
+              "inference_time_ms": 0,
+              "tool_time_ms": 1155,
+              "active_time_ms": 1155
+            }
+          }
+        },
+        "next_node_id": "implement",
+        "git_commit_sha": "dec9d74c1c37868bb5e51d2ffa3a5690c4784e5f",
+        "node_visits": {
+          "preflight_lint": 1,
+          "preflight_compile": 1,
+          "start": 1,
+          "toolchain": 1
+        }
+      },
+      "diff": {
+        "summary": {
+          "files_changed": 0,
+          "additions": 0,
+          "deletions": 0
+        }
+      }
+    },
+    {
+      "seq": 0,
+      "checkpoint": {
+        "timestamp": "2026-07-01T16:44:45.145968107Z",
+        "current_node": "implement",
+        "completed_nodes": [
+          "start",
+          "toolchain",
+          "preflight_compile",
+          "preflight_lint",
+          "implement"
+        ],
+        "node_retries": {},
         "context_values": {
           "thread.preflight_compile.current_node": "preflight_lint",
           "internal.node_visit_count": 1,
           "internal.retry_count.toolchain": 0,
+          "last_response": "Implemented Plan A in `fabro-redact`.\n\nSummary:\n- Added `SecretRedactor` in `lib/crates/fabro-redact/src/secret_registry.rs`.\n- Exported it from `lib/crates/fabro-redact/src/lib.rs`.\n- Reused the crat",
+          "response.implement": "Implemented Plan A in `fabro-redact`.\n\nSummary:\n- Added `SecretRedactor` in `lib/crates/fabro-redact/src/secret_registry.rs`.\n- Exported it from `lib/crates/fabro-redact/src/lib.rs`.\n- Reused the crate’s shared `REDACTION_MARKER` constant.\n- Implemented:\n  - shared cloneable per-run state via `Arc<RwLock<Vec<String>>>`\n  - `register`, ignoring empty/whitespace-only values and de-duplicating\n  - `redact_into`, using longest-first exact-match redaction\n  - `redact_json`, redacting all JSON string leaves while preserving object keys\n  - `is_empty`\n- Added unit tests covering low-entropy values, empty values, overlaps, empty identity behavior, nested JSON, and shared clone state.\n\nTDD:\n- Red step confirmed: `cargo nextest run -p fabro-redact` failed before implementation because `SecretRedactor` was missing.\n- Green step passed after implementation.\n\nVerification run:\n- `cargo +nightly-2026-04-14 fmt --check --all`\n- `cargo +nightly-2026-04-14 clippy --workspace --all-targets -- -D warnings`\n- `cargo nextest run -p fabro-redact`\n- `cargo dev build -- -p fabro-cli --release`\n- `cargo dev docs check`\n- `cd apps/fabro-web && bun run typecheck`\n- `cd lib/packages/fabro-api-client && bun run typecheck`\n- `cargo nextest run --workspace`\n\nAll passed. Note: the first release build attempt failed because Bun workspace dependencies were not installed; I ran `bun install --frozen-lockfile` and reran successfully. No tracked files were changed by that install.",
           "internal.retry_count.start": 0,
           "graph.goal": "# Plan A — `SecretRedactor` in `fabro-redact`\n\n**This is Plan A of three** (split for parallel execution):\n\n- **Plan A (this file)** — add a per-run secret-value redactor to `fabro-redact`.\n  Self-contained; touches only `fabro-redact/`. **Run in parallel with Plan B.**\n- **Plan B** — resolve `secrets.*` tokens at the run boundary. **Run in parallel\n  with Plan A.**\n- **Plan C** — wire redaction across leak surfaces + hooks. **Run after A and B\n  merge** (it consumes this crate's type and Plan B's lookup).\n\nThis plan is inert on its own: it adds a tested library primitive that Plan C\nwires up. Shipping it alone changes no behavior.\n\n> **Token notation.** Interpolation tokens are written in this file without their\n> enclosing double curly braces, so the file is safe to pass directly as a\n> workflow goal (the goal templater would otherwise try to expand them). Read\n> `secrets.NAME`, `env.NAME`, and `secrets.*` as the double-curly-brace-wrapped\n> token form used everywhere else in the codebase, and write the real\n> double-brace syntax in the code, tests, and docs you produce.\n\n---\n\n## Overall goal (shared context)\n\nMake secret tokens (`secrets.NAME`) in workflow config resolve from the server\nvault, at the run boundary, with values that never get persisted, never leak, and\nfail closed when a secret is missing or the wrong type. The redaction guarantee\nfor declared secrets is: content-based redaction (already present) is the\nuniversal baseline, plus a per-run registry of resolved secret **values** so a\ndeclared secret is redacted even when it does not look like a credential. **This\nplan builds that registry primitive.**\n\nWhy per-run and not a process global: a test-only in-process run path executes\nmultiple runs in the same process, so redaction state must be per-run, never a\n`static`/global.\n\n## Conventions\n\n- **TDD.** Write the failing test first, then the code.\n- Match the codebase: Rust import style (types by name, functions via parent\n  module, no glob imports in production), `strum` for enum string maps, keep\n  test-only helpers behind `#[cfg(test)]`.\n- Plain-English commit messages, PR text, and comments — no internal planning\n  identifiers.\n- The verify gate runs nightly `fmt --check`, nightly\n  `clippy --all-targets -D warnings`, `cargo nextest run --workspace`, docs check,\n  web/api-client typecheck, and a release build. Implement so all pass.\n- Never print or log a secret value.\n\n---\n\n## Implementation\n\n### A.1 — Add the `SecretRedactor` type\n\nFile: new `lib/crates/fabro-redact/src/secret_registry.rs`, exported from\n`lib/crates/fabro-redact/src/lib.rs`.\n\nAdd a cheap, cloneable, per-run registry of secret values that redacts exact\nmatches regardless of shape. It composes *after* the existing content-based\nredaction (`redact_string`, `redact_json_value`) — this type does not replace\nthem.\n\nShape:\n\n- `SecretRedactor` backed by shared, interior-mutable state (e.g.\n  `Arc<Mutex<Vec<String>>>` or `Arc<RwLock<...>>`) so a clone handed to a\n  different subsystem observes registrations. Derive `Clone` and `Default`; an\n  empty redactor is a pure no-op.\n- `fn register(&self, value: impl Into<String>)` — store a secret value to be\n  redacted. **Ignore empty or whitespace-only values** (registering an empty\n  string would turn all output into `REDACTED`). De-duplicate.\n- `fn redact_into(&self, s: &str) -> String` — replace every registered value\n  substring with the same `\"REDACTED\"` marker used by `redact_string`. Replace\n  **longest values first** so a secret that is a substring of another is handled\n  correctly. If the registry is empty, return the input unchanged (fast path).\n- `fn redact_json(&self, value: serde_json::Value) -> serde_json::Value` — walk\n  the JSON tree and apply `redact_into` to every string leaf (both object values\n  and array elements; object keys are left as-is). Exact-value matching is\n  unambiguous, so unlike `redact_json_value` this pass does not skip any keys.\n- Optional `fn is_empty(&self) -> bool` for callers that want to skip work.\n\nReuse the crate's existing `\"REDACTED\"` replacement marker (see `redact_string`\nin `lib.rs`) rather than introducing a new literal.\n\n### A.2 — Tests (unit, in the new module)\n\n- A **low-entropy** value (e.g. `\"staging\"`) that `redact_string` would *not*\n  catch is replaced with `REDACTED` by `redact_into` after `register(\"staging\")`.\n- Registering `\"\"` or `\"   \"` is a no-op: `redact_into` leaves unrelated text\n  intact (guard against the empty-value footgun).\n- Overlapping values: register both `\"abc\"` and `\"abcdef\"`; `redact_into` on a\n  string containing `\"abcdef\"` redacts the whole token (longest-first), not just\n  the `\"abc\"` prefix.\n- Empty registry: `redact_into` and `redact_json` are the identity.\n- `redact_json` redacts a registered value nested inside an object value and\n  inside an array element.\n- A clone of the redactor observes values registered through the original (shared\n  state), proving it can be handed to another subsystem.\n\n### A.3 — Verify\n\n- `cargo +nightly-2026-04-14 fmt --check --all`\n- `cargo +nightly-2026-04-14 clippy --workspace --all-targets -- -D warnings`\n- `cargo nextest run -p fabro-redact`\n- release build (`cargo dev build -- -p fabro-cli --release`)\n\n## Dependencies\n\nNone. Parallel-safe with Plan B. This type is consumed by Plan C.\n",
           "failure_class": "",
           "outcome": "succeeded",
           "command.output": "blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126",
           "graph.rankdir": "LR",
+          "thread.preflight_lint.current_node": "implement",
           "internal.run_id": "01KWF7MM3VPXZZA8BTHJXE9VT1",
-          "internal.thread_id": "preflight_compile",
+          "internal.thread_id": "preflight_lint",
           "failure_signature": "",
           "thread.start.current_node": "toolchain",
           "internal.retry_count.preflight_compile": 0,
-          "current_node": "preflight_lint",
+          "current_node": "implement",
           "internal.retry_count.preflight_lint": 0,
           "internal.work_dir": "/home/daytona/workspace/fabro",
           "internal.fidelity": "compact",
           "thread.toolchain.current_node": "preflight_compile",
+          "internal.retry_count.implement": 0,
+          "last_stage": "implement",
           "graph.model_stylesheet": "\n            * { model: claude-opus-4-8; }\n        "
         },
         "node_outcomes": {
@@ -768,12 +873,49 @@
               "tool_time_ms": 153109,
               "active_time_ms": 153109
             }
+          },
+          "implement": {
+            "status": "succeeded",
+            "context_updates": {
+              "last_stage": "implement",
+              "last_response": "Implemented Plan A in `fabro-redact`.\n\nSummary:\n- Added `SecretRedactor` in `lib/crates/fabro-redact/src/secret_registry.rs`.\n- Exported it from `lib/crates/fabro-redact/src/lib.rs`.\n- Reused the crat",
+              "response.implement": "Implemented Plan A in `fabro-redact`.\n\nSummary:\n- Added `SecretRedactor` in `lib/crates/fabro-redact/src/secret_registry.rs`.\n- Exported it from `lib/crates/fabro-redact/src/lib.rs`.\n- Reused the crate’s shared `REDACTION_MARKER` constant.\n- Implemented:\n  - shared cloneable per-run state via `Arc<RwLock<Vec<String>>>`\n  - `register`, ignoring empty/whitespace-only values and de-duplicating\n  - `redact_into`, using longest-first exact-match redaction\n  - `redact_json`, redacting all JSON string leaves while preserving object keys\n  - `is_empty`\n- Added unit tests covering low-entropy values, empty values, overlaps, empty identity behavior, nested JSON, and shared clone state.\n\nTDD:\n- Red step confirmed: `cargo nextest run -p fabro-redact` failed before implementation because `SecretRedactor` was missing.\n- Green step passed after implementation.\n\nVerification run:\n- `cargo +nightly-2026-04-14 fmt --check --all`\n- `cargo +nightly-2026-04-14 clippy --workspace --all-targets -- -D warnings`\n- `cargo nextest run -p fabro-redact`\n- `cargo dev build -- -p fabro-cli --release`\n- `cargo dev docs check`\n- `cd apps/fabro-web && bun run typecheck`\n- `cd lib/packages/fabro-api-client && bun run typecheck`\n- `cargo nextest run --workspace`\n\nAll passed. Note: the first release build attempt failed because Bun workspace dependencies were not installed; I ran `bun install --frozen-lockfile` and reran successfully. No tracked files were changed by that install."
+            },
+            "notes": "Stage completed: implement",
+            "usage": {
+              "input": {
+                "usage": {
+                  "model": {
+                    "provider": "openai",
+                    "model_id": "gpt-5.5"
+                  },
+                  "tokens": {
+                    "input_tokens": 463942,
+                    "output_tokens": 6369,
+                    "reasoning_tokens": 6173,
+                    "cache_read_tokens": 791552,
+                    "cache_write_tokens": 0
+                  }
+                },
+                "facts": {
+                  "algorithm": "openai"
+                }
+              },
+              "total_usd_micros": 3091746
+            },
+            "timing": {
+              "wall_time_ms": 0,
+              "inference_time_ms": 549060,
+              "tool_time_ms": 652102,
+              "active_time_ms": 1201162
+            }
           }
         },
-        "next_node_id": "implement",
+        "next_node_id": "simplify_opus",
         "node_visits": {
           "start": 1,
           "preflight_compile": 1,
+          "implement": 1,
           "preflight_lint": 1,
           "toolchain": 1
         }
@@ -807,6 +949,207 @@
   "superseded_by": null,
   "pending_interviews": {},
   "stages": {
+    "implement@1": {
+      "first_event_seq": 52,
+      "prompt": null,
+      "response": null,
+      "completion": null,
+      "provider_used": {
+        "mode": "agent",
+        "provider": "openai",
+        "model": "gpt-5.5",
+        "reasoning_effort": "xhigh"
+      },
+      "diff": null,
+      "script_invocation": null,
+      "script_timing": null,
+      "parallel_results": null,
+      "output": null,
+      "started_at": "2026-07-01T16:24:43.676345229Z",
+      "handler": "agent",
+      "usage": {
+        "input_tokens": 463942,
+        "output_tokens": 6369,
+        "total_tokens": 1268036,
+        "reasoning_tokens": 6173,
+        "cache_read_tokens": 791552,
+        "cache_write_tokens": 0,
+        "total_usd_micros": 3091746
+      },
+      "model": {
+        "provider": "openai",
+        "model_id": "gpt-5.5"
+      },
+      "permission_level": "full",
+      "agent_tools": [
+        {
+          "name": "apply_patch",
+          "description": "Use the `apply_patch` tool to edit files. This is a FREEFORM tool, so do not wrap the patch in JSON.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "write",
+          "invoked": true
+        },
+        {
+          "name": "close_agent",
+          "description": "Close a running subagent that is no longer needed.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "subagent",
+          "invoked": false
+        },
+        {
+          "name": "glob",
+          "description": "Find files by file names using a glob pattern. Use path to choose the search root. Prefer this over shell find or ls when locating repository files.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "read",
+          "invoked": true
+        },
+        {
+          "name": "grep",
+          "description": "Search file contents with a regex pattern. Use path to choose the search root, glob_filter to limit matching files, case_insensitive for case folding, and max_results to cap output.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "read",
+          "invoked": true
+        },
+        {
+          "name": "read_file",
+          "description": "Read files before editing them. Returns line-numbered text and supports offset/limit for large files. Use this instead of shell cat, head, tail, or sed when inspecting repository files.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "read",
+          "invoked": true
+        },
+        {
+          "name": "request_user_input",
+          "description": "Ask the human one or more questions and wait for their answers before continuing this stage.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "other",
+          "invoked": false
+        },
+        {
+          "name": "send_input",
+          "description": "Send a follow-up message to a running subagent when new information or corrected instructions are needed.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "subagent",
+          "invoked": false
+        },
+        {
+          "name": "shell",
+          "description": "Execute shell commands for terminal operations, package managers, tests and builds. Use dedicated tools for file reads, file edits, filename searches, and content searches. Provide timeout_ms for long-running commands.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "shell",
+          "invoked": true
+        },
+        {
+          "name": "spawn_agent",
+          "description": "Spawn a subagent for independent work or context isolation. Use it for tasks that can proceed separately, and avoid duplicating the same work in the parent session.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "subagent",
+          "invoked": false
+        },
+        {
+          "name": "update_plan",
+          "description": "Update the multi-step plan for the current task. Submit the entire plan; existing steps are reconciled by exact step text.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "other",
+          "invoked": false
+        },
+        {
+          "name": "wait",
+          "description": "Wait for a subagent to complete, then use the result to synthesize the outcome for the user.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "subagent",
+          "invoked": false
+        },
+        {
+          "name": "web_fetch",
+          "description": "Fetch content from a URL that starts with http:// or https://. Pass a prompt to extract specific information or summarize the page; omit prompt to return the page content.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "other",
+          "invoked": false
+        },
+        {
+          "name": "web_search",
+          "description": "Search the web using Brave Search when current external information is needed. Returns result titles, URLs, and descriptions; use web_fetch for a specific URL.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "other",
+          "invoked": false
+        },
+        {
+          "name": "write_file",
+          "description": "Create new files, or overwrite an existing file only when replacement is explicitly intended. Prefer edit_file for targeted changes to existing files because write_file overwrites the full file content.",
+          "source": {
+            "kind": "native"
+          },
+          "category": "write",
+          "invoked": false
+        }
+      ],
+      "context_window": {
+        "provider": "openai",
+        "model": "gpt-5.5",
+        "context_window_tokens": 272000,
+        "input_tokens": 58100,
+        "usage_percent": 21.360294117647058,
+        "count_method": "response_usage_scaled_breakdown",
+        "staleness": "live",
+        "generated_at": "2026-07-01T16:44:45.126524162Z",
+        "event_seq": 202,
+        "breakdown": [
+          {
+            "category": "system_prompt",
+            "tokens": 984,
+            "usage_percent": 0.36176470588235293
+          },
+          {
+            "category": "tools",
+            "tokens": 1403,
+            "usage_percent": 0.5158088235294118
+          },
+          {
+            "category": "memory",
+            "tokens": 3339,
+            "usage_percent": 1.2275735294117647
+          },
+          {
+            "category": "conversation",
+            "tokens": 52368,
+            "usage_percent": 19.25294117647059
+          },
+          {
+            "category": "other",
+            "tokens": 6,
+            "usage_percent": 0.0022058823529411764
+          }
+        ],
+        "warnings": []
+      },
+      "state": "running"
+    },
     "start@1": {
       "first_event_seq": 18,
       "prompt": null,
@@ -941,7 +1284,12 @@
       "first_event_seq": 42,
       "prompt": null,
       "response": null,
-      "completion": null,
+      "completion": {
+        "outcome": "succeeded",
+        "notes": "Script completed: cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1",
+        "failure_reason": null,
+        "timestamp": "2026-07-01T16:24:40.578472462Z"
+      },
       "provider_used": null,
       "diff": null,
       "script_invocation": {
@@ -949,11 +1297,27 @@
         "command": "exec 2>&1\ncargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1",
         "language": "shell"
       },
-      "script_timing": null,
+      "script_timing": {
+        "output": "blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126",
+        "exit_code": 0,
+        "duration_ms": 153109,
+        "termination": "exited",
+        "output_bytes": 0,
+        "live_streaming": false
+      },
       "parallel_results": null,
       "output": null,
+      "output_bytes": 0,
+      "live_streaming": false,
+      "termination": "exited",
       "started_at": "2026-07-01T16:22:07.463678393Z",
       "handler": "command",
+      "timing": {
+        "wall_time_ms": 153114,
+        "inference_time_ms": 0,
+        "tool_time_ms": 153109,
+        "active_time_ms": 153109
+      },
       "usage": {
         "input_tokens": 0,
         "output_tokens": 0,
@@ -962,7 +1326,7 @@
         "cache_read_tokens": 0,
         "cache_write_tokens": 0
       },
-      "state": "running"
+      "state": "succeeded"
     }
   }
 }
\ No newline at end of file
diff --git a/stages/004-preflight_lint@1/output.log b/stages/004-preflight_lint@1/output.log
new file mode 100644
index 000000000..d87ba9545
--- /dev/null
+++ b/stages/004-preflight_lint@1/output.log
@@ -0,0 +1 @@
+blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126
\ No newline at end of file
diff --git a/stages/004-preflight_lint@1/script_timing.json b/stages/004-preflight_lint@1/script_timing.json
new file mode 100644
index 000000000..db502f0a7
--- /dev/null
+++ b/stages/004-preflight_lint@1/script_timing.json
@@ -0,0 +1,8 @@
+{
+  "output": "blob://sha256/12ae32cb1ec02d01eda3581b127c1fee3b0dc53572ed6baf239721a03d82e126",
+  "exit_code": 0,
+  "duration_ms": 153109,
+  "termination": "exited",
+  "output_bytes": 0,
+  "live_streaming": false
+}
\ No newline at end of file
diff --git a/stages/004-preflight_lint@1/status.json b/stages/004-preflight_lint@1/status.json
new file mode 100644
index 000000000..9532bc8c7
--- /dev/null
+++ b/stages/004-preflight_lint@1/status.json
@@ -0,0 +1,6 @@
+{
+  "outcome": "succeeded",
+  "notes": "Script completed: cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1",
+  "failure_reason": null,
+  "timestamp": "2026-07-01T16:24:40.578472462Z"
+}
\ No newline at end of file
diff --git a/stages/005-implement@1/prompt.md b/stages/005-implement@1/prompt.md
new file mode 100644
index 000000000..26ef0cb1a
--- /dev/null
+++ b/stages/005-implement@1/prompt.md
@@ -0,0 +1,129 @@
+Goal: # Plan A — `SecretRedactor` in `fabro-redact`
+
+**This is Plan A of three** (split for parallel execution):
+
+- **Plan A (this file)** — add a per-run secret-value redactor to `fabro-redact`.
+  Self-contained; touches only `fabro-redact/`. **Run in parallel with Plan B.**
+- **Plan B** — resolve `secrets.*` tokens at the run boundary. **Run in parallel
+  with Plan A.**
+- **Plan C** — wire redaction across leak surfaces + hooks. **Run after A and B
+  merge** (it consumes this crate's type and Plan B's lookup).
+
+This plan is inert on its own: it adds a tested library primitive that Plan C
+wires up. Shipping it alone changes no behavior.
+
+> **Token notation.** Interpolation tokens are written in this file without their
+> enclosing double curly braces, so the file is safe to pass directly as a
+> workflow goal (the goal templater would otherwise try to expand them). Read
+> `secrets.NAME`, `env.NAME`, and `secrets.*` as the double-curly-brace-wrapped
+> token form used everywhere else in the codebase, and write the real
+> double-brace syntax in the code, tests, and docs you produce.
+
+---
+
+## Overall goal (shared context)
+
+Make secret tokens (`secrets.NAME`) in workflow config resolve from the server
+vault, at the run boundary, with values that never get persisted, never leak, and
+fail closed when a secret is missing or the wrong type. The redaction guarantee
+for declared secrets is: content-based redaction (already present) is the
+universal baseline, plus a per-run registry of resolved secret **values** so a
+declared secret is redacted even when it does not look like a credential. **This
+plan builds that registry primitive.**
+
+Why per-run and not a process global: a test-only in-process run path executes
+multiple runs in the same process, so redaction state must be per-run, never a
+`static`/global.
+
+## Conventions
+
+- **TDD.** Write the failing test first, then the code.
+- Match the codebase: Rust import style (types by name, functions via parent
+  module, no glob imports in production), `strum` for enum string maps, keep
+  test-only helpers behind `#[cfg(test)]`.
+- Plain-English commit messages, PR text, and comments — no internal planning
+  identifiers.
+- The verify gate runs nightly `fmt --check`, nightly
+  `clippy --all-targets -D warnings`, `cargo nextest run --workspace`, docs check,
+  web/api-client typecheck, and a release build. Implement so all pass.
+- Never print or log a secret value.
+
+---
+
+## Implementation
+
+### A.1 — Add the `SecretRedactor` type
+
+File: new `lib/crates/fabro-redact/src/secret_registry.rs`, exported from
+`lib/crates/fabro-redact/src/lib.rs`.
+
+Add a cheap, cloneable, per-run registry of secret values that redacts exact
+matches regardless of shape. It composes *after* the existing content-based
+redaction (`redact_string`, `redact_json_value`) — this type does not replace
+them.
+
+Shape:
+
+- `SecretRedactor` backed by shared, interior-mutable state (e.g.
+  `Arc<Mutex<Vec<String>>>` or `Arc<RwLock<...>>`) so a clone handed to a
+  different subsystem observes registrations. Derive `Clone` and `Default`; an
+  empty redactor is a pure no-op.
+- `fn register(&self, value: impl Into<String>)` — store a secret value to be
+  redacted. **Ignore empty or whitespace-only values** (registering an empty
+  string would turn all output into `REDACTED`). De-duplicate.
+- `fn redact_into(&self, s: &str) -> String` — replace every registered value
+  substring with the same `"REDACTED"` marker used by `redact_string`. Replace
+  **longest values first** so a secret that is a substring of another is handled
+  correctly. If the registry is empty, return the input unchanged (fast path).
+- `fn redact_json(&self, value: serde_json::Value) -> serde_json::Value` — walk
+  the JSON tree and apply `redact_into` to every string leaf (both object values
+  and array elements; object keys are left as-is). Exact-value matching is
+  unambiguous, so unlike `redact_json_value` this pass does not skip any keys.
+- Optional `fn is_empty(&self) -> bool` for callers that want to skip work.
+
+Reuse the crate's existing `"REDACTED"` replacement marker (see `redact_string`
+in `lib.rs`) rather than introducing a new literal.
+
+### A.2 — Tests (unit, in the new module)
+
+- A **low-entropy** value (e.g. `"staging"`) that `redact_string` would *not*
+  catch is replaced with `REDACTED` by `redact_into` after `register("staging")`.
+- Registering `""` or `"   "` is a no-op: `redact_into` leaves unrelated text
+  intact (guard against the empty-value footgun).
+- Overlapping values: register both `"abc"` and `"abcdef"`; `redact_into` on a
+  string containing `"abcdef"` redacts the whole token (longest-first), not just
+  the `"abc"` prefix.
+- Empty registry: `redact_into` and `redact_json` are the identity.
+- `redact_json` redacts a registered value nested inside an object value and
+  inside an array element.
+- A clone of the redactor observes values registered through the original (shared
+  state), proving it can be handed to another subsystem.
+
+### A.3 — Verify
+
+- `cargo +nightly-2026-04-14 fmt --check --all`
+- `cargo +nightly-2026-04-14 clippy --workspace --all-targets -- -D warnings`
+- `cargo nextest run -p fabro-redact`
+- release build (`cargo dev build -- -p fabro-cli --release`)
+
+## Dependencies
+
+None. Parallel-safe with Plan B. This type is consumed by Plan C.
+
+
+## Completed stages
+- **toolchain**: succeeded
+  - Script: `command -v cargo >/dev/null || { curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y && sudo ln -sf $HOME/.cargo/bin/* /usr/local/bin/; }; cargo --version 2>&1`
+  - Output:
+    ```
+    cargo 1.95.0 (f2d3ce0bd 2026-03-21)
+    ```
+- **preflight_compile**: succeeded
+  - Script: `cargo check -q --workspace 2>&1`
+  - Output: (empty)
+- **preflight_lint**: succeeded
+  - Script: `cargo +nightly-2026-04-14 clippy -q --workspace --all-targets -- -D warnings 2>&1`
+  - Output: (empty)
+
+
+Read the plan file referenced in the goal and implement every step. Make all the code changes described in the plan. Use red/green TDD.
\ No newline at end of file
diff --git a/stages/005-implement@1/provider_used.json b/stages/005-implement@1/provider_used.json
new file mode 100644
index 000000000..c57772db6
--- /dev/null
+++ b/stages/005-implement@1/provider_used.json
@@ -0,0 +1,6 @@
+{
+  "mode": "agent",
+  "provider": "openai",
+  "model": "gpt-5.5",
+  "reasoning_effort": "xhigh"
+}
\ No newline at end of file
diff --git a/stages/005-implement@1/response.md b/stages/005-implement@1/response.md
new file mode 100644
index 000000000..9207d11b0
--- /dev/null
+++ b/stages/005-implement@1/response.md
@@ -0,0 +1,29 @@
+Implemented Plan A in `fabro-redact`.
+
+Summary:
+- Added `SecretRedactor` in `lib/crates/fabro-redact/src/secret_registry.rs`.
+- Exported it from `lib/crates/fabro-redact/src/lib.rs`.
+- Reused the crate’s shared `REDACTION_MARKER` constant.
+- Implemented:
+  - shared cloneable per-run state via `Arc<RwLock<Vec<String>>>`
+  - `register`, ignoring empty/whitespace-only values and de-duplicating
+  - `redact_into`, using longest-first exact-match redaction
+  - `redact_json`, redacting all JSON string leaves while preserving object keys
+  - `is_empty`
+- Added unit tests covering low-entropy values, empty values, overlaps, empty identity behavior, nested JSON, and shared clone state.
+
+TDD:
+- Red step confirmed: `cargo nextest run -p fabro-redact` failed before implementation because `SecretRedactor` was missing.
+- Green step passed after implementation.
+
+Verification run:
+- `cargo +nightly-2026-04-14 fmt --check --all`
+- `cargo +nightly-2026-04-14 clippy --workspace --all-targets -- -D warnings`
+- `cargo nextest run -p fabro-redact`
+- `cargo dev build -- -p fabro-cli --release`
+- `cargo dev docs check`
+- `cd apps/fabro-web && bun run typecheck`
+- `cd lib/packages/fabro-api-client && bun run typecheck`
+- `cargo nextest run --workspace`
+
+All passed. Note: the first release build attempt failed because Bun workspace dependencies were not installed; I ran `bun install --frozen-lockfile` and reran successfully. No tracked files were changed by that install.
\ No newline at end of file