What it is
security-guidance is an official plugin in Anthropic's Claude Code plugin directory. It has three layers. On every Edit or Write it checks about 25 known-dangerous patterns, such as yaml.load, pickle.load on untrusted data, raw innerHTML and hardcoded secrets, with no model call. When Claude finishes a turn, a separate model call reviews the diff and sends high-severity findings back to Claude to fix. On git commit, an agentic reviewer reads related files to trace data across the codebase, catching problems such as insecure direct object references, auth bypass and cross-file SSRF.
Teams can add plain-language rules in .claude/claude-security-guidance.md. The star count belongs to the whole plugin directory, not this plugin; the directory has no tagged releases.
Our review
When the same AI writes the code and checks it, it tends to agree with itself. security-guidance gets around that by sending each turn's diff to a fresh model call with no memory of why the code was written.
It works at three depths. Every edit gets an instant check against about 25 risky patterns, such as pickle and innerHTML,, which costs nothing. Every finished turn gets a model review of the diff. Every commit gets a deeper agent that reads surrounding files, which is how it finds bugs that span several files, such as one user reading another user's records.
The catch: none of this blocks anything. Findings go back to Claude as instructions, and Claude can still miss or ignore them. The two model-backed reviews cost usage on every turn and commit, and the plugin's own docs suggest pairing it with a blocking hook or a CI check. The star count you see is the whole plugin directory's, not this plugin's: about 44,000 at the time of writing. It needs Claude Code 2.1.144 or later and Python 3.8+.
Our take: think of it as a colleague reading over your shoulder, not a lock on the door. It is worth having for the price of one install command, as long as something like Semgrep in CI still stands between the code and production.
Where it earns its place
A two-person startup shipping a web app with Claude Code
Catch injection and auth bugs before they reach main
/plugin install security-guidance@claude-plugins-official. Pattern checks are free; the turn and commit reviews add model usage to each session.
An AppSec lead at a company rolling out Claude Code
Encode in-house security rules once for every developer
Commit .claude/claude-security-guidance.md with the rules in plain language, and keep a blocking CI scanner as the actual gate.
Before you commit
- Setup
- Run /plugin install security-guidance@claude-plugins-official inside Claude Code; needs Python 3.8+ on PATH.
- Integrations
- Claude Code hooks; reviews run through your Claude sign-in, an API key, or Bedrock, Vertex or Foundry.
- Data and privacy
- Diffs and related code go to the Claude model you use for reviews, under that account's terms. Pattern checks run locally.
- Cost at scale
- Free plugin; the per-turn diff review and per-commit agentic review run on the newest Opus by default, configurable with SECURITY_REVIEW_MODEL, and bill as normal usage.
- Limits
- Non-blocking by design; review quality depends on the model and can miss issues.
What's good
- One install command, no configuration
- Separate model call reviews each diff with fresh context
- Commit review traces data across files
- Custom rules in plain language
What's not
- Advises only; never blocks a write or commit
- Model-backed reviews cost usage on every turn
- Needs Python on PATH
- Stars reflect the whole plugin directory
Best for
Claude Code users who want security review on every change without setting anything up.
Alternatives
security-guidance plugin (Claude Code): common questions
- Is security-guidance plugin (Claude Code) worth it?
- For a Claude Code user this is a one-command security win, with a reviewer that is not grading its own homework. We rate it 4.2 out of 5.
- How much does security-guidance plugin (Claude Code) cost?
- Free plugin under Apache 2.0; the two model-backed reviews use your Claude or API usage.
- What are the downsides of security-guidance plugin (Claude Code)?
- Advises only; never blocks a write or commit; model-backed reviews cost usage on every turn; needs Python on PATH; stars reflect the whole plugin directory.
- Who is security-guidance plugin (Claude Code) best for?
- Claude Code users who want security review on every change without setting anything up.
- How much does security-guidance plugin (Claude Code) cost for a team or in production?
- Free plugin; the per-turn diff review and per-commit agentic review run on the newest Opus by default, configurable with SECURITY_REVIEW_MODEL, and bill as normal usage.
- Is security-guidance plugin (Claude Code) safe for company data?
- Diffs and related code go to the Claude model you use for reviews, under that account's terms. Pattern checks run locally.
- What are the best security-guidance plugin (Claude Code) alternatives?
- The closest alternatives we have reviewed are Trail of Bits Skills and Cloudflare Security Audit Skill.
Answers checked Oct 24, 2026.
Make security-guidance plugin (Claude Code)? Show this rating on your site
