The Agent Skills Directory

[SAFE]: The skill is a meta-utility for reviewing and optimizing other skills. It includes 'Hard Gates' and 'Red Flags' designed to enforce security and quality best practices, such as requiring explicit stop conditions for high-risk side effects (e.g., payments or deletions).
[COMMAND_EXECUTION]: The instructions mandate the execution of local project scripts, including scripts/generate-index.sh, scripts/generate-router.sh, and scripts/update-lock.sh. These tools are used for internal metadata synchronization and repository management within the local project context.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it is designed to process and analyze external skill definitions.
Ingestion points: The skill reads target SKILL.md files and auxiliary documentation identified in the 'Entry/Exit' section.
Boundary markers: The skill does not prescribe technical delimiters for ingested content, relying on the auditor's instructions and 'Iron Laws' to maintain control during the review.
Capability inventory: The skill possesses capabilities to read files, write audit reports to the docs/ directory, and execute local maintenance scripts.
Sanitization: Content validation is performed semantically against the 'Iron Law' and 'Verification Checklist' rather than through automated string sanitization.

verify-team-skill-quality