The Agent Skills Directory

[SAFE]: The skill consists primarily of structured documentation, templates, and mathematical models for evaluation. It does not contain executable scripts or binary files, providing only illustrative code examples in markdown.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to review and score external artifacts (code and docs), which involves processing untrusted data. However, this is the skill's primary intended purpose. The recommended verification tools (such as pytest, radon, jscpd, and lychee) are standard development utilities. The risk is minimized by the skill's focus on systematic, metric-based review.
[EXTERNAL_DOWNLOADS]: The documentation suggests installing well-known Python packages (e.g., textstat, pytest) from official registries for quality measurement. These references to established tools from trusted ecosystems do not pose a security threat.

evaluation-framework