6a63e35e75
Fixes anthropics/claude-plugins-official#2056 — on Windows, when the worktree contains an untracked file whose name has a character undefined in cp1252 (accented capitals like Á Í Ï Ð Ý, most CJK, emoji), the UserPromptSubmit hook crashes: Exception in thread Thread-5 (_readerthread): UnicodeDecodeError: 'charmap' codec can't decode byte 0x81 Traceback (most recent call last): File diffstate.py, line 338, in _list_untracked for p in r.stdout.split('\\0'): AttributeError: 'NoneType' object has no attribute 'split' Non-blocking (UPS failures still let the prompt through) but the baseline-untracked snapshot is silently lost, so the Stop-hook review mis-handles pre-existing untracked files. Root cause (reporter's diagnosis, verified): 1. core.quotePath=false makes git emit raw UTF-8 for non-ASCII filenames. 2. subprocess.run(..., text=True) decodes via locale.getpreferredencoding(False) in strict mode — on Windows that is cp1252, in which 0x81 / 0x8D / 0x8F / 0x90 / 0x9D are undefined. Those bytes appear in the UTF-8 encodings of Á (C3 81), Í (C3 8D), Ï (C3 8F), Ð (C3 90), Ý (C3 9D), and a large fraction of CJK / emoji codepoints. 3. The decode runs in the subprocess reader thread. The thread raises UnicodeDecodeError, threading prints 'Exception in thread Thread-N', subprocess.run returns with stdout=None. The handler then does None.split('\\0') -> AttributeError, which is NOT in the narrow except (TimeoutExpired, FileNotFoundError, OSError) tuple, so it escapes the helper, propagates out of UserPromptSubmit's ThreadPoolExecutor.result(), and exits the hook non-zero. This is internally inconsistent: gitutil._git_diff_range, security_reminder_hook._reflog_amend_lookup (line ~540), and the commit diff loop (line ~1115) already do bytes + decode utf-8/replace, with comments explicitly noting that text=True would crash. The fix below extends that established pattern to the helpers that were holdouts. Affected helpers (6 total): - diffstate._list_untracked <- reporter, hot path, CRITICAL - diffstate.capture_git_baseline <- reporter, latent - diffstate.get_baseline_file_content <- audit, file content read, HIGH - gitutil._git_name_only <- reporter, latent - gitutil._git_status_porcelain <- reporter, latent - gitutil._git_reflog_recent_commits <- audit, embeds %gs commit msg, HIGH For each one: - Drop text=True from subprocess.run. - Decode r.stdout / r.stderr as .decode('utf-8', errors='replace'). - Add ValueError to the except tuple as defense against any future strict-decode regression (UnicodeDecodeError is a ValueError subclass; including it explicitly degrades the helper to its empty/None return instead of escaping out of the hook). Verified locally on macOS Python 3.13: - py_compile clean on both files. - 45 existing smoke + extensibility tests still pass. - 21 new internal tests (not in this PR — added to the team's local test suite at staging/tests/test_unicode_decode.py): * 18 static-shape parametrized: each of the 6 fixed helpers has no text=True in its subprocess calls, contains errors='replace', and lists ValueError in its except. * Deterministic end-to-end: create real git repo + Ávila_report.txt untracked, call _list_untracked, verify it returns {'Ávila_report.txt': <mtime>} without crashing. * Deterministic end-to-end: same for capture_git_baseline (verifies the latent stderr-warning case stays valid). * Deterministic end-to-end: get_baseline_file_content on a file whose content has 山田太郎 + 🎉; verify the bytes round-trip through the decode. - 66/66 tests pass total (45 existing + 21 new). NOT verified end-to-end on Windows — would need actual cp1252 strict decode to fire. Reporter has the deterministic repro and will re-verify on their Win11 / Python 3.14.x setup before merge. Not in this PR (defense-in-depth, lower risk): - 3 git rev-parse calls returning path output (gitutil._find_git_index, _git_toplevel, _git_dir) could fail on Windows if cwd is in a non-ASCII install directory. Same fix shape but unreported and much lower probability — worth a separate follow-up if anyone actually hits it. 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