Skip to main content
skill CX Codex

vuln-check — Claude Code skill: what it does, how to install

vuln-check is a Claude Code skill that checks resolved dependencies and related code for security vulnerabilities. It runs for security requests or /vuln-check.

Apache-2.0 ★ 515 updated 2026-11-17

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add liujh0729-jpg/AI-enterprise-OEM --skill vuln-check -a claude-code
Open on GitHub

What vuln-check does

vuln-check reviews resolved dependencies and the code paths connected to them for security issues. It reads app/pubspec.yaml, packages/*/pubspec.yaml, sdk/plugin-kit/package.json, and the matching lockfiles. Advisory data is used to record the resolved version, affected range, advisory, and fixed version when available, while keeping an advisory match separate from an exploit demonstrated in the project.

The skill follows untrusted input into shell execution, filesystem access, plugin capabilities, WebView or HTML injection, and bridge HTTP routes. Path checks cover containment and symlink behavior. Potential secrets are reported by path, line, and type only. Findings include severity, location, trigger, impact, evidence, and the smallest effective remedy.

A check does not run automatic dependency fixes. When advisory data cannot be obtained, the dependency review is incomplete, and an empty result is not treated as proof that the project has no vulnerabilities.

What it helps with

  • check resolved dependencies against current advisories
  • trace untrusted input into shell and filesystem operations
  • inspect plugin, WebView, and bridge HTTP boundaries
  • scan committed files for secrets without exposing values
  • separate confirmed issues from leads that need verification

Example prompts for the agent

  • Run /vuln-check across the project and report dependency and code-path vulnerabilities.
  • Trace user-controlled input into shell execution, filesystem access, and bridge HTTP routes.
  • Perform a security check and list potential secrets by path, line, and type only.

How to install vuln-check

1Skills CLI (Vercel)

npx skills add liujh0729-jpg/AI-enterprise-OEM --skill vuln-check -a claude-code

2Manually: put the files in place

~/.claude/skills/vuln-check/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/vuln-check/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about vuln-check

How do I install vuln-check in Claude Code?

Run npx skills add liujh0729-jpg/AI-enterprise-OEM --skill vuln-check -a claude-code, then use ~/.claude/skills/vuln-check/SKILL.md.

What does vuln-check do?

It reviews resolved dependencies and relevant code paths, follows untrusted input into sensitive operations, and produces findings with severity, location, trigger, impact, and evidence.

How does vuln-check handle filesystem paths?

It checks path containment and symlink behavior. It also inspects actual validation and authorization before treating a suspicious API as a vulnerability.

Source text

SKILL.md · Apache-2.0
---
name: vuln-check
description: Check resolved dependencies and relevant code paths for security vulnerabilities. Use for security checks or /vuln-check; use audit for a broader correctness and maintainability assessment.
---

Assess the requested security surface with evidence. A check returns findings; implement remediation when the user has requested it.

1. Read `app/pubspec.yaml`, `packages/*/pubspec.yaml`, `sdk/plugin-kit/package.json` and the relevant lockfiles for the affected dependency paths. Use current advisory data for the ecosystem involved, and record the resolved version, advisory, affected range, and verified fixed version where available. Distinguish a dependency advisory match from a demonstrated exploit path in this project.
2. Trace untrusted inputs through guards to sensitive operations: shell execution, filesystem access, plugin capabilities, WebView/HTML injection, and bridge HTTP routes. For paths, check containment and symlink behavior where relevant; normalization alone does not prevent traversal. Inspect actual validation and authorization before treating a suspicious API as a vulnerability.
   Scan committed repository files for potential secrets; report only the path, line, and secret type, never the value or a source excerpt.
3. Report findings by severity with location, trigger, impact, evidence, and the smallest effective remedy. Separate confirmed issues from leads needing verification. Preserve the upstream peer/dev dependency contract when proposing upgrades.
4. State the scope, sources checked, and gaps. If advisories are unavailable, report that the dependency check is incomplete; absence of findings is not a claim that the project is vulnerability-free.

Report potential secrets only by path, line, and type, never their value or a source excerpt. Do not run automatic dependency fixes as part of a check; for requested remediation, make targeted changes and validate the affected paths.

© liujh0729-jpg · Apache-2.0 · liujh0729-jpg/AI-enterprise-OEM

Similar skills

All skills for Security →

From the same set3

Project sponsors

Companies that keep this analytics open