Review AI agent skill source, code, and permissions to classify risk levels.
npx skills add https://github.com/fishyer/skills --skill skill-vetter-fishyer
Please help me install this Agent Skill. Skill: skill-vetter Source: https://github.com/fishyer/skills/tree/main/skills/skill-vetter Command: npx skills add https://github.com/fishyer/skills --skill skill-vetter-fishyer
This Skill provides a security-first approach to vetting AI agent skills before installation, identifying potential risks and suspicious patterns.
Run 'skill-vetter on a new skill to automatically check for security risks.
High-intent search queries and answers about installing and using this skill.
To check an AI agent skill for security risks, you can run a vetting protocol that reviews the source code, validates author reputation, and evaluates file and network permission scopes to prevent unauthorized data access.
Skill vetting is a security-first review process that identifies malicious code patterns and classifies permissions. It categorizes assessed skills into low, medium, high, or extreme risk levels based on detected red flags.
You identify malicious code patterns by conducting an automated code review that scans for suspicious red flags, evaluates network access scope, and checks the source reputation before installing the AI agent skill.
Yes, the skill vetting process explicitly evaluates the scope of file and network access. This permission scope assessment helps prevent potential security breaches and unauthorized data access by installed AI agent skills.
You need to conduct a risk assessment on AI agent skills before installation. This security-first vetting protocol prevents potential security breaches by validating the source and identifying suspicious code patterns early.
j4hr3nAudit AI agent permissions, tools, paths, and credentials for security risks.
mi179Create isolated Cloudflare Workers sandboxes for executing untrusted code.
DevHive1Orchestrate dependency, static analysis, and privacy dataflow scanners into prioritized reports.
gl0bal01Execute malware samples in sandboxes and capture runtime behavior with monitoring tools.
