What does the Security Guidance plugin do?

It reviews Claude-generated code with pattern-based warnings on edits, an LLM-powered diff review on Stop, and an agentic commit reviewer that catches injection, XSS, SSRF, hardcoded secrets, and 25+ other vulnerability classes.

Which vulnerability classes does it detect?

It catches injection, XSS, SSRF, and hardcoded secrets, along with 25+ other vulnerability classes.

When does the review run?

Pattern-based warnings appear on edits as they happen, the LLM-powered diff review runs on Stop, and the agentic commit reviewer examines commits.

security-guidance

PackSecurity

Security Guidance is a Claude plugin that reviews code written by AI agents. It flags risky patterns as edits happen, runs an LLM-powered diff review when a task stops, and uses an agentic commit reviewer to catch injection, XSS, SSRF, hardcoded secrets, and 25+ other vulnerability classes. The security guidance Claude plugin gives developers a way to check agent-generated code before it lands.

Unavailable. Delivery for this kind is on the roadmap — not serving yet.

Have a setup where Claude plugins can be installed.

What your AI can do with it

Pattern-based warnings on code edits as they happen
LLM-powered diff review triggered on Stop
Agentic commit reviewer that inspects commits
Detects injection, XSS, SSRF, and hardcoded secrets
Covers 25+ vulnerability classes

Getting started

Have a setup where Claude plugins can be installed.
Add the Security Guidance plugin.
Let the agent make code edits so pattern-based warnings appear.
Review the LLM diff review output when a task stops.
Check the agentic commit reviewer's findings before committing.

Signals

GitHub stars: 37k
Forks: 4k
Last commit: Sep 2026

Questions

What does the Security Guidance plugin do?: It reviews Claude-generated code with pattern-based warnings on edits, an LLM-powered diff review on Stop, and an agentic commit reviewer that catches injection, XSS, SSRF, hardcoded secrets, and 25+ other vulnerability classes.
Which vulnerability classes does it detect?: It catches injection, XSS, SSRF, and hardcoded secrets, along with 25+ other vulnerability classes.
When does the review run?: Pattern-based warnings appear on edits as they happen, the LLM-powered diff review runs on Stop, and the agentic commit reviewer examines commits.

Advanced

Item type: plugin
Key: anthropics-claude-plugins-official-security-guidance
Source: github.com/anthropics/claude-plugins-official

github.com/anthropics/claude-plugins-official