bc07f7a1fd
URGENT REGRESSION FIX. PR #2076 (Graphite gt workflow) gated the PostToolUse commit/push hooks with: "if": "Bash(git commit:*)|Bash(gt create:*)|Bash(gt modify:*)" "if": "Bash(git push:*)|Bash(gt submit:*)" mirroring the regex-OR idiom that `matcher` uses ("Edit|Write|MultiEdit|NotebookEdit"). But `if` is NOT a regex — it's a SINGLE permission-rule string. The CC harness's dispatch filter parses the entire `if` value as one rule of shape `ToolName(rule_content)` via: let firstParen = H.indexOf("("); let lastParen = H.lastIndexOf(")"); // searches from END if (lastParen !== H.length - 1) return { toolName: H }; let toolName = H.slice(0, firstParen); let ruleContent = H.slice(firstParen + 1, lastParen); Applied to the broken commit clause: toolName = "Bash" ruleContent = "git commit:*)|Bash(gt create:*)|Bash(gt modify:*" The garbled `ruleContent` never matches any real command, so the hook never fires — for ANY workflow, not just gt. The plugin's deepest review layer was dead in production for all users on builds shipping PR #2076. Fix shape: split into separate hook entries, each with its own well-formed single-rule `if` clause. The Python hook self-routes commit vs push via the bash-command regexes and dedups concurrent spawns via `_claim_bash_hook_once`, so multiple entries firing the same script is safe. This commit: 1. hooks.json: 5 precise entries (one per command shape) instead of the broken |-joined 2-entry form. Restores the original commit/ push behavior bit-for-bit (`Bash(git commit:*)` + `Bash(git push:*)` are unchanged from pre-#2076), and adds 3 separate entries for the Graphite commands (`Bash(gt create:*)`, `Bash(gt modify:*)`, `Bash(gt submit:*)`). No git behavior change. The earlier draft used the broader `Bash(git *)` + `Bash(gt *)` per the reporter's suggestion, but that has a real cost: every `git status` / `git log` / `git diff` would spawn the Python hook only to early-exit via the regex matcher. Precise per-command entries avoid the spawn overhead and match the pre-#2076 cost profile exactly. 2. security_reminder_hook.py: widen `_GIT_COMMIT_RE` to tolerate `git -C <path>` and `git -c k=v` global options between `git` and `commit` (mirrors `_GIT_PUSH_RE`'s long-standing tolerance). Without this, `git -C /repo commit` is silently dropped by the handler — reporter flagged this as the secondary finding. Verified locally on macOS Python 3.13: - hooks.json valid JSON, 5 `if` clauses each parses to a single `{toolName: "Bash", ruleContent: "<command>:*"}` pair. - py_compile security_reminder_hook.py clean. - 9-case regex sanity: all 4 commit forms match (bare, -C path, -c k=v, gt create/modify); 3 non-commit forms reject (status, gt submit, gt log). Pre-fix would reject -C path form. - 7 new tests in test_2089_if_clause_validity.py + 2 updated tests in test_gt_graphite_workflow.py: * 12 sanity tests for a Python parser mirroring harness's BA(H) — pinned so a future refactor can't silently start accepting the broken form. * 2 hooks.json validity: every `if` clause parses as a single valid rule; at least one if-gated hook exists. * 1 post-fix structure: separate entries cover git AND gt. * 2 updated gt-coverage: SOME clause covers git, SOME clause covers gt (no longer requires both in the same |-joined clause, which was the broken shape). TDD-verified the test catches the bug: temporarily restored main's broken |-joined hooks.json, ran the new test, saw `test_every_if_clause_is_single_valid_rule` fail with a clear error explaining #2089's cause. Restored fix, test passes. - Full suite: 336/353 pass (17 unrelated failures from open PRs #2078 / #2086 not in this branch). NOT verified end-to-end with a real CC instance triggering the hooks on a git or gt commit. The static-shape tests catch the regression class and the regex sanity tests pin the `git -C` tolerance, but the asyncRewake feedback loop needs runtime verification. Closes #2089. Restores the closes for #2048 that PR #2076 attempted. 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