70c28b9c2f
Fixes #2159. The Stop hook emits feedback via `hookSpecificOutput: {hookEventName: "Stop", additionalContext}`, but `Stop`/`SubagentStop` are NOT members of CC's `hookSpecificOutput` discriminated union (coreSchemas.ts — valid members are PreToolUse, PostToolUse, UserPromptSubmit, SessionStart, etc.). So the emitted shape violates CC's documented hook-output schema. Impact is CC-version-dependent — important nuance, established empirically: - Reporter (array0224-cloud) on CLI 2.1.150 / 2.1.152: CC rejects the Stop feedback; the block/reason never reaches the model, so the auto-rewake/fix loop is lost. (Detection still runs + logs.) - On CLI 2.1.160 (current) the asyncRewake completion path is lenient: its gate is `isSyncHookJSONOutput` (hooks.ts) which is just `!(json.async === true)` — NOT a strict schema parse. So the invalid hookSpecificOutput is tolerated: metrics + rewakeSummary are still consumed and the model still receives the findings. I could NOT reproduce the rejection on 2.1.160, and BQ confirms Stop-path vulns_found metrics are recorded normally (~21k with-vuln fires / 3d), i.e. NOT dropped. (An earlier draft of this message claimed metrics were dropped — that was wrong; corrected after checking telemetry + repro'ing the old plugin on 2.1.160.) So this is defensive schema-correctness: the plugin should emit output that conforms to CC's documented union regardless of how strictly a given CC version validates it. The reporter's environment validates strictly; relying on the current version's leniency is fragile. Fix (CC's documented asyncRewake "clean pattern" — hooks.ts: "error text on stderr, JSON on stdout"): - For Stop/SubagentStop, emit_metrics writes guidance to stderr (the asyncRewake body channel CC delivers via `stderr || stdout`) and sets top-level `decision: "block"` + `reason` (valid SyncHookJSONOutput fields; also the documented sync Stop-hook contract for the `-p` fallback). It does NOT emit a Stop hookSpecificOutput. - PostToolUse (commit-review, push-sweep) is unchanged — valid union member, keeps the modern hookSpecificOutput protocol. Verified locally on macOS Python 3.13: - py_compile clean. - 11 new tests (test_2159_stop_hook_schema.py) pin the contract: Stop output carries no hookSpecificOutput, uses top-level decision/reason, writes guidance to stderr; the emitted hookEventName (when present) is a valid union member. 2 existing tests that asserted the buggy Stop->hookSpecificOutput shape were corrected. Full suite 464/464 pass + 2 skipped. - END-TO-END in /tmux on CLI 2.1.160: * FIXED plugin (2.0.3): staged pickle.loads + os.system, benign edit pulls the file into review_set; Stop LLM review found 2 critical vulns; CC delivered a clean rewake ("Background security review found issues" + both findings). All hooks (UPS, PostToolUse[Edit] pattern, PostToolUse[Bash] commit-review, Stop) fired clean; zero schema rejections / errors / http_err in the debug log. * OLD plugin (2.0.2) on the SAME 2.1.160: also delivered Stop feedback (confirming the no-repro-on-latest finding above) — which proves the fix carries NO regression risk on current CC while making the output robust for the stricter versions where it actually breaks. Version bumped 2.0.2 -> 2.0.3 per the per-PR-bump policy (#2114: a bump is the only way the fix reaches the existing fleet — relevant for users still on the CC versions where this breaks). Closes #2159. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
security-guidance
Security review for Claude-generated code. Three layers:
- Pattern warnings — instant regex-based reminders on
Edit/Writefor ~25 known-dangerous patterns (yaml.load,torch.load(weights_only=False),pickle.loadon untrusted data, rawinnerHTML, hardcoded secrets, etc.). - LLM diff review — when Claude finishes a turn, the plugin sends the diff to a fast LLM call (Opus 4.7 by default) and feeds high-severity findings back to Claude so it can fix them before you see the response.
- Agentic commit review — on
git commit, an SDK-driven reviewer reads related files (Read/Grep/Glob) to trace data flow across the codebase, catching multi-file vulnerabilities pattern matching misses (IDOR, auth bypass, cross-file SSRF).
Findings cover common web-vulnerability classes — injection, XSS, SSRF, hardcoded secrets, IDOR, auth bypass, unsafe deserialization, and path traversal among others.
Install
/plugin install security-guidance@claude-plugins-official
Marketplace ships enabled by default in Claude Code — no setup beyond having the CLI itself.
Prerequisites
- Claude Code CLI ≥ v2.1.144
- Python 3.8+ on
PATH(python3,python, orpy -3— the plugin picks the first that works) - A working API path (subscription, API key, or 3P provider config)
Configuration
All configuration is via environment variables. None are required for default behavior.
Selecting a model
# 1P / gateway: a canonical model id
SECURITY_REVIEW_MODEL=claude-opus-4-7 # default
# Bedrock: use the inference-profile id
SECURITY_REVIEW_MODEL=us.anthropic.claude-opus-4-7
# Vertex: use the Vertex date-tag form
SECURITY_REVIEW_MODEL=claude-opus-4-7@20260218
SECURITY_REVIEW_MODEL controls the LLM diff review. SG_AGENTIC_MODEL (same syntax) controls the agentic commit reviewer; defaults to the same model.
Enabling/disabling layers
| Variable | Default | What it does |
|---|---|---|
SECURITY_GUIDANCE_DISABLE=1 |
unset | Kill switch — disables the entire plugin |
ENABLE_PATTERN_RULES=0 |
on | Disable layer 1 (regex pattern warnings) |
ENABLE_CODE_SECURITY_REVIEW=0 |
on | Disable all LLM reviews (Stop hook + commit/push) |
ENABLE_STOP_REVIEW=0 |
on | Disable only the Stop-hook diff review, keeping commit/push reviews. Useful for multi-agent / shared-worktree setups where another agent can move HEAD between a worker's turns |
ENABLE_COMMIT_REVIEW=0 |
on | Disable layer 3 (agentic commit review) |
Higher-recall mode
SG_DUAL_OR=on # default off
Runs two parallel review calls and unions the findings. Catches a few percentage points more vulnerabilities in our testing, at roughly 2× the API cost per review. Most users don't need it.
Org-specific policies
Drop a claude-security-guidance.md in any of:
~/.claude/claude-security-guidance.md— user-wide rules<project>/.claude/claude-security-guidance.md— project rules, intended to be committed<project>/.claude/claude-security-guidance.local.md— local overrides, intended to be.gitignore'd
All three are loaded and concatenated into the LLM diff review's prompt in the order user → project → project-local. If the combined size exceeds the 8 KB prompt budget, the tail is truncated, so user-wide rules are kept and project-local rules are dropped first. The agentic commit reviewer (layer 3) does not currently read this file. Example:
# Acme security rules
- All SELECTs against the `customers` or `orders` tables MUST go through `db.replica`,
never `db.primary`. Primary is for writes only.
- Background jobs must not use the user-context auth token; they get
service-account creds from `jobs.get_service_account()`.
- Calls to `requests.get(url)` with a user-controlled `url` need
the SSRF-allowlist wrapper at `acme.net.safe_request`.
Built-in rules cover common web-vulnerability classes without it — claude-security-guidance.md is for things specific to your codebase that the model can't infer.
Privacy and data handling
The plugin sends data to a model endpoint to perform its reviews. Specifically, each Stop-hook diff review transmits the changed file paths, the diff hunks, and the relevant file contents in the diff; each agentic commit review additionally transmits any files the reviewer pulls in via Read/Grep/Glob while tracing data flow. Your claude-security-guidance.md contents (user, project, and local) are appended to the prompt on every review, so don't put secrets in it.
Where that data goes depends on your Claude Code configuration:
- Default (Anthropic API / subscription): sent to
api.anthropic.comand handled under Anthropic's Commercial Terms and Privacy Policy. - LLM gateway (
ANTHROPIC_BASE_URLset): sent to your gateway URL instead. The gateway operator's terms apply. - 3rd-party providers (Bedrock / Vertex / Foundry / Mantle): sent to your configured provider endpoint. The provider's data-handling terms apply (e.g., AWS / GCP / Azure).
The plugin writes its own debug log to ~/.claude/security/log.txt (override with SECURITY_GUIDANCE_DEBUG_LOG). The log contains diffstate metadata and finding categories — no full file contents or model prompts — and rotates at 1 MB. Nothing is uploaded.
Limitations
This is a best-effort assistive tool, not a guarantee. Treat findings as suggestions, not as a substitute for human code review, SAST/DAST, dependency scanning, or pen-testing. The reviewer can miss vulnerabilities, produce false positives, and may behave differently across codebases, languages, and model versions. No warranty is provided — use is subject to Anthropic's Commercial Terms.
Troubleshooting
Plugin doesn't seem to fire — check that ~/.claude/claude-security-guidance.md (or hook activity) shows in debug logs. Run Claude Code with --debug-file /tmp/claude/debug.txt and grep for security_reminder_hook. The plugin also writes its own log to ~/.claude/security/log.txt.
Review never finds anything — verify your API path works. On 3P providers, check SECURITY_REVIEW_MODEL is set to a provider-specific id (not a bare claude-opus-4-7). On LLM gateways, check the gateway's logs for POST /v1/messages traffic from the plugin.
Too many false positives — drop SECURITY_REVIEW_MODEL to a cheaper model (claude-sonnet-4-6) and re-evaluate; if precision is the priority, stay on Opus 4.7.
Want to silence a specific finding — add a comment to the line explaining why it's safe; the LLM reviewer treats inline justifications as exclusions. For systemic exclusions, document them in your claude-security-guidance.md.
Reporting issues
Open an issue on the security-guidance plugin repo with:
- The Claude Code CLI version (
claude --version) - Provider setup (1P / Bedrock / Vertex / LLM gateway / etc.)
- A minimal repro diff
- The relevant section of
~/.claude/security/log.txt