Audit AI agent skills for malicious patterns and prompt injection risks.
npx skills add https://github.com/gula00/autoclaw-skills --skill skill-vetter-gula00
Please help me install this Agent Skill. Skill: skill-vetter Source: https://github.com/gula00/autoclaw-skills/tree/main/skill-vetter-1.0.0 Command: npx skills add https://github.com/gula00/autoclaw-skills --skill skill-vetter-gula00
This Skill provides a crucial security layer by vetting AI agent skills before they are installed, preventing the introduction of malicious code, prompt injection, or data exfiltration risks.
Use the skill-vetter to analyze the skill located at the GitHub URL 'https://github.com/example/new-skill'.
High-intent search queries and answers about installing and using this skill.
To perform a security audit on an AI agent skill, use this tool to analyze its source code for malicious patterns, unauthorized data exfiltration, and prompt injection vulnerabilities. It classifies identified risks and provides a clear installation verdict.
The best way to check a GitHub repository for prompt injection risks is to run a pre-installation security audit. This process analyzes the skill's source code to identify malicious patterns and classifies them into Low, Medium, High, or Extreme risk categories.
Yes, you can vet AI agent skills for unauthorized data exfiltration before installing them. The vetting process specifically analyzes source code for unauthorized network access and credential exposure to prevent data theft.
Risk assessment classifies malicious code patterns in AI skills by analyzing the source code for suspicious behavior and potential impact. It categorizes the threat into Low, Medium, High, or Extreme risk levels to determine installation safety.
The limitations of automated security vetting for AI skills include potential false positives or missing novel obfuscation techniques in source code. It provides a risk classification and installation verdict but cannot guarantee absolute safety from all zero-day vulnerabilities.
KhoiwallFetch source code from npm, PyPI, crates.io, and GitHub/GitLab repositories.
damianpapadopoulosAssess autonomous agent architectural safety for lethal-trifecta risks.
DevHive1Orchestrate dependency, static analysis, and privacy dataflow scanners into prioritized reports.
RuifengFuReview AI agent skills for malicious patterns and permission risks.
