5212308979
Fixes anthropics/claude-plugins-official#2048 — teams using Graphite for stacked PRs (`gt create` / `gt modify` / `gt submit`) never get the commit/push agentic review because the hook matcher only catches literal `git commit` / `git push` Bash calls. gt shells out to git as a subprocess, but the hook fires on Claude's top-level tool call, which is `gt create` — not the `git commit` invocation inside the gt subprocess that Claude Code never observes. Per-edit pattern checks and end-of-turn Stop review still fire (those don't depend on detecting the commit command), so the silent-coverage- gap is bounded to the deepest review layer for Graphite users. Still: that's exactly the layer designed to catch IDOR / auth-bypass / cross-file SSRF, so the gap matters. Semantic mapping (per the reporter): gt create -> commit (like `git commit`) gt modify -> commit + amend (like `git commit --amend`) gt submit -> push (like `git push`) Changes: 1. hooks/hooks.json: extend two PostToolUse `if` matchers. "Bash(git commit:*)" -> "Bash(git commit:*)|Bash(gt create:*)|Bash(gt modify:*)" "Bash(git push:*)" -> "Bash(git push:*)|Bash(gt submit:*)" Without this, the hook subprocess never spawns for gt invocations and the Python regex changes below are dead code. 2. hooks/security_reminder_hook.py: extend three regexes that classify the bash command line. _GIT_COMMIT_RE: now also matches `gt create` and `gt modify`. Used at 4 sites (handler gate, multi-commit count, prompt detection, event classification). Compound commands like `gt create -am a && gt submit` now correctly trigger both the commit and push paths. _GIT_AMEND_RE: now also matches `gt modify` (semantically an amend). The amend code path uses reflog to find the pre-amend SHA and diff against THAT instead of HEAD~1 — same code path now applies to `gt modify`. _GIT_PUSH_RE: now also matches `gt submit`. Tolerates the same `git -C path` / `git -c k=v` global options as before for the git form; gt has its own flag layer that doesn't conflict. Verified locally on macOS Python 3.13: - JSON valid (hooks.json roundtrips). - Existing 45 smoke + extensibility tests still pass. - 76 new tests in test_gt_graphite_workflow.py (added to internal test suite this PR doesn't ship — kept in sg-staging tests/ until we have a story for shipping plugin tests publicly): * 16 parametrized commit-match: native git commit variants + all gt create / gt modify variants from the reporter's repro. * 11 parametrized commit-reject: gt submit, gt log, gtoolkit (word-boundary), agt create, etc. * 9 parametrized amend-match: git commit --amend variants + gt modify variants + chained git+gt. * 7 parametrized amend-reject: regular git commit, gt create, gt submit, echo'd substring noise. * 11 parametrized push-match: git push variants + gt submit variants + chained. * 12 parametrized push-reject: git commit, gt log, gt fetch, gt down, gt restack, gh pr create, agt submit. * 3 compound-command class tests: git+gt mixtures trigger both paths; gt modify chained with gt submit triggers amend + push. * 3 commit-invocation-count tests: gt commands contribute to the multi-commit-detection findall count. * 2 hooks.json static config tests: read the JSON, verify the commit and push `if` clauses include the gt cases. Catches the easy regression where someone updates the Python regex but forgets to widen the matcher. - 121/121 pass total (45 existing + 76 new) in 2.50s. NOT verified end-to-end with a real `gt` install. Reporter has the deterministic Graphite workflow and offered to retest. The regex + matcher widening is a clean superset — current git-only matching still works (verified by the 45-test smoke suite that uses `git commit` / `git push` exclusively), and the new gt cases are pure additions. Not in this PR: - `gt prev` / `gt next` / `gt up` / `gt down` etc. — pure navigation, no commit / push side effect. - `gt restack` — could in principle rewrite commits (so the plugin's reviewed-shas cache becomes stale), but it doesn't create reviewable new content. Out of scope. - `gh pr create` — already explicitly NOT a separate matcher per the existing comment in _GIT_PUSH_RE (gh invokes git push as a child process; the bash hook only sees the top-level `gh pr create`). Same architectural issue as gt but with a different cost/benefit per the existing comment. 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