Skill Security Auditor — Security & Safety

Skill Security Auditor is a security audit and vulnerability scanner for AI agent skills before installation. It scans skill directories or git repos for malicious code, prompt injection, dependency risks, and file system boundary violations, producing a PASS/WARN/FAIL verdict with findings and remediation guidance.

engineering claude-code

📂 View skill on GitHub ↗ 🚀 How to use

✨ What it does

Scans code files for command injection, code execution, obfuscation, network exfiltration, credential harvesting, file s
Detects prompt injection patterns in SKILL.md and reference markdown files.
Checks dependencies for known vulnerabilities, typosquatting, unpinned versions, and install commands in code.
Verifies file system access stays within skill boundaries and flags hidden files, binaries, large files, and symlinks.
Provides a clear PASS/WARN/FAIL verdict with severity-grouped findings and remediation guidance.
Supports strict mode, JSON output, git repo auditing, CI/CD integration, and batch audits.

🎯 When to use it

Evaluating a skill from an untrusted source before installation.
Auditing a skill directory or git repo URL for malicious code.
Pre-install security gate for Claude Code plugins, OpenClaw skills, or Codex skills.
Scanning Python scripts for dangerous patterns like os.system, eval, subprocess, network exfiltration.
Detecting prompt injection in SKILL.md files or checking dependency supply chain risks.

🚀 How to use

Trigger the skill by asking to audit a skill, e.g., 'audit this skill', 'is this skill safe', or 'scan skill for security'. Provide a local skill directory path or a git repo URL (optionally with --skill to specify a subdirectory). The skill runs a Python script that outputs a report; use --strict to treat warnings as failures, --json for JSON output, and --cleanup to remove temporary clones.

Example prompts:

audit this skill /path/to/skill-name/
scan skill for security https://github.com/user/repo --skill skill-name

📄 Output: A security audit report with a PASS/WARN/FAIL verdict, findings grouped by severity, and remediation guidance.

📦 Add this skill to Claude Code

# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills

# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/engineering/skills/skill-security-auditor .claude/skills/skill-security-auditor

Skill source: engineering/skills/skill-security-auditor/SKILL.md

⚠️ Good to know

Static analysis only; cannot detect logic bombs or time-delayed payloads with certainty, and obfuscation detection is pattern-based so a sufficiently creative attacker may bypass it.

❓ FAQ

What verdicts can the audit produce?

The audit produces PASS (no critical or high findings), WARN (high/medium findings requiring manual review), or FAIL (critical findings, do not install without remediation).

Does the skill execute the code it scans?

No, it performs static analysis only, which is safe but less complete than dynamic analysis.

Can I audit a skill from a git repository before cloning?

Yes, you can pass a git repo URL and optionally use --cleanup to clone to a temp directory, audit, and then clean up.

🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.