Does it support auditing skills from remote repositories?

Yes, you can provide a Git repository URL, and the auditor will analyze the remote skill, including optional cleanup of temporary files after the scan.

What types of files does the auditor scan?

The auditor scans Python (.py), Shell (.sh, .bash), JavaScript/TypeScript (.js, .ts) files, as well as SKILL.md files and dependency manifests like requirements.txt and package.json.

Can it detect all possible security threats?

It uses advanced static analysis to identify known malicious patterns and risks. However, as it does not execute the code, it may not detect complex logic bombs or time-delayed payloads.

How should I interpret the audit verdicts?

PASS indicates the skill is safe for installation. WARN suggests high/medium risks that require manual review. FAIL identifies critical security violations that should not be installed.

What kind of prompt injection risks does it look for?

It scans for system prompt overrides, role hijacking (e.g., 'Act as root'), safety bypasses, and hidden instructions inside markdown files that could manipulate the AI's behavior.

Skill Security Auditor

Name: Skill Security Auditor
Author: zhangzhang-111-i

byzhangzhang-111-i

0•

보안 및 테스팅

Audits and scans AI agent skills for security vulnerabilities, malicious code, and prompt injection risks prior to installation.

Skill Security Auditor acts as a critical security gate for the AI agent ecosystem, specifically designed to protect environments from malicious plugins and untrusted code. It performs rigorous static analysis on Python, shell scripts, and markdown files to detect command injection, data exfiltration, credential harvesting, and suspicious network activity. By providing a clear PASS/WARN/FAIL verdict along with actionable remediation guidance, this skill helps developers safely evaluate third-party skills from Git repositories or local directories, ensuring they remain within safe operational boundaries.

주요 기능

01Dependency supply chain risk assessment for PyPI and npm packages

02Prompt injection detection for SKILL.md and documentation files

03Filesystem boundary verification to prevent unauthorized directory access

04Automated PASS/WARN/FAIL reporting with detailed remediation steps

05Multi-language static analysis for Python, Bash, and JavaScript/TypeScript scripts

060 GitHub stars

사용 사례

01Detecting hidden data exfiltration patterns in untrusted Python utility scripts

02Vetting third-party GitHub repositories before installing a new Claude Code skill

03Integrating automated security gates into CI/CD pipelines for AI skill development

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add zhangzhang-111-i/claude-skills111 skill-security-auditor

For use in Claude.ai and ChatGPT

주요 기능

01Dependency supply chain risk assessment for PyPI and npm packages

02Prompt injection detection for SKILL.md and documentation files

03Filesystem boundary verification to prevent unauthorized directory access

04Automated PASS/WARN/FAIL reporting with detailed remediation steps

05Multi-language static analysis for Python, Bash, and JavaScript/TypeScript scripts

060 GitHub stars

사용 사례

01Detecting hidden data exfiltration patterns in untrusted Python utility scripts

02Vetting third-party GitHub repositories before installing a new Claude Code skill

03Integrating automated security gates into CI/CD pipelines for AI skill development

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add zhangzhang-111-i/claude-skills111 skill-security-auditor

For use in Claude.ai and ChatGPT