Add a local Claude review worker for essential pull request heads

docs/github-automation.md

@@ -71,6 +71,19 @@ The Actions allow-list below was checked against live settings and pinned source
   removed: the API refused every request that carried the action's attribution, on every stored Console key, while
   the same keys work from the CLI and the Agent SDK. Reviews now come from the local Claude review worker, the GPT
   and trading reads, and Codex's GitHub review; the required checks and the landing scripts enforce the gates.
+- **Local review of essential heads** ([`tools/claude-review-worker/`](../tools/claude-review-worker/README.md), added
+  2026-10-09): a command-center-owned systemd user timer reviews, every 15 minutes, this repository's same-repository
+  heads that touch `blueprints/us-equities/`, `.github/`, `scripts/validate*.py`, `tools/credentials/`,
+  `adoption/hooks/` or `tools/local-pages/`, and every head of the private trading repository. Claude Code headless runs
+  on a stored API key inside a bubblewrap sandbox with no home directory (no gh login, no key store), the key passed by
+  `tools/credentials/credential_run.py` through the environment only; Read, Glob and Grep only, no CLAUDE.md memory
+  (`CLAUDE_CODE_DISABLE_CLAUDE_MDS=1`), a $10 client budget, an 11 USD bound per run and a 55 USD daily ceiling in the
+  api-actions ledger. A trading head's prompt also carries the owner's upstream-alignment rule: each claim about
+  external behaviour must cite an upstream pin or a vendor URL. A model-free step posts the commit
+  status `claude-review/local` (here the status only, its description built from parsed fields; on the private
+  repository also one sanitized comment), which the command center's landing scripts can require on essential pull
+  requests; the ruleset does not. Posting stays off until the command center installs it and sets
+  `CLAUDE_REVIEW_POST=1`, and no real review has run ([decision](decisions/2026-10-09-claude-review-worker.md)).
 - **Against the final catalog of 2026-10-01** (`docs/final-catalog-20261001.md`, #595): the picks each blind model
   family made for these layers. The clean-room definitive round announced there decides one pick per slot.
   - git-github-automation: both families picked git, gh, Worktrunk and sem (the Claude pick keeps sem only if a

manifests/evidence.json

@@ -18625,6 +18625,11 @@
       "sha256": "98947f11f2a3eb693c9c31b3fff6e58e1f20d476411aa49f42f66888d43c6d47",
       "bytes": 54090
     },
+    {
+      "path": "docs/decisions/2026-10-09-claude-review-worker.md",
+      "sha256": "8ac125de3ef17e149391e08175b00f67cf365d0ebcd8a5bb0a91f301d2e3337c",
+      "bytes": 62373
+    },
     {
       "path": "docs/decisions/2026-10-09-codex-native-review-trial.md",
       "sha256": "337ec932d1e70994adc7b6e90458114e604d70ec065c49fc7d7ff8c2191e11fc",
@@ -18872,8 +18877,8 @@
     },
     {
       "path": "docs/github-automation.md",
-      "sha256": "5b12ec4c5b5323d1ff39987a710de18cab5e2c1beb916c5a6918e076a42643d1",
-      "bytes": 102555
+      "sha256": "b3622e272e6cbfb083ebbb40e853f0451de066e8fcab2db9a2f91c07b5753f19",
+      "bytes": 104021
     },
     {
       "path": "docs/grand-catalog-counts.md",
@@ -61455,6 +61460,11 @@
       "sha256": "bc84e4527cbbd894ad524981ee5c90230552a42cb53e7088699ee9a0d2730093",
       "bytes": 11326
     },
+    {
+      "path": "tests/test_claude_review_worker.py",
+      "sha256": "890d644757305e86a3157c325f05ddeca2588ece27faa8d39f642b2901aa7691",
+      "bytes": 130675
+    },
     {
       "path": "tests/test_claude_spend_scan.py",
       "sha256": "bd4994e38cdadfa5b4099351261c6908875540d5be54139aa9e4f8dae4691e4b",
@@ -62745,6 +62755,36 @@
       "sha256": "3efdc4b54ba33b11c67f8f2449167de03b558e8c53c7e09cdb5a9f66593311a7",
       "bytes": 22476
     },
+    {
+      "path": "tools/claude-review-worker/README.md",
+      "sha256": "aaccf9237e8ad77968e452c0f6bfbfcc6ea23d77374d0aa0776ae9b8a891d4c5",
+      "bytes": 13873
+    },
+    {
+      "path": "tools/claude-review-worker/essential-paths.json",
+      "sha256": "68c9ea2f3f50d996da35491ff3a0d1892bb217b67006ca2fd07eae094030a9f3",
+      "bytes": 4190
+    },
+    {
+      "path": "tools/claude-review-worker/probes.py",
+      "sha256": "e8c6b5b00da9bd1011835a42a9768e31a427ce34b137ea9676d726010d311bf6",
+      "bytes": 51641
+    },
+    {
+      "path": "tools/claude-review-worker/systemd/claude-review-worker.service",
+      "sha256": "b6b5268e4a4a0cdb91ce351cf2c8554a02fe2c466fe2740d7bad15d33207be18",
+      "bytes": 2680
+    },
+    {
+      "path": "tools/claude-review-worker/systemd/claude-review-worker.timer",
+      "sha256": "b2a31d4cdadecf261f1a38a6ead9f170898a59013c900be4d9d86f4f7cf78503",
+      "bytes": 648
+    },
+    {
+      "path": "tools/claude-review-worker/worker.py",
+      "sha256": "30d5fbd68f6c521a3336a104df7b1cec223177ff6c3008d66cdfc7d7fef270a6",
+      "bytes": 129379
+    },
     {
       "path": "tools/credentials/canary_lane_consumer.sh",
       "sha256": "e820394825eeea9a1cfcec3c2ec87ccb62400af28c929f6ab2d52a99cd0a810c",