agent-skill-evaluator

To check an agent skill for prompt injection risks, provide the GitHub URL to initiate a security assessment that inspects provenance, SKILL.md content, scripts, and references, producing a risk score with concrete remediation guidance.

Agent skill security evaluation is a pre-flight safety check that verifies creator legitimacy and inspects skill definitions for safety signals. You need it before installing any external skill to prevent web scraping vulnerabilities or malicious code execution.

Assess the safety of a .skill URL by initiating a comprehensive evaluation that aggregates community feedback, verifies source credibility, and generates a structured security assessment document with evidence, severity ratings, and a final verdict.

Yes, you can get an automated risk score without manual code review. The evaluation automatically inspects scripts, assets, and SKILL.md definitions to generate a final risk verdict with actionable remediation steps.

Yes, the skill evaluator detects web scraping and external reference vulnerabilities by inspecting references and scripts for safety signals, returning a structured document with severity ratings and evidence for any identified risks.