The Agent Skills Directory

[SAFE]: The skill provides comprehensive guidance for creating and refactoring AI agent skills. It does not contain any malicious instructions or hidden payloads.
[COMMAND_EXECUTION]: The skill includes a local utility script scripts/validate_skill_dir.ts (written for the Bun runtime) that performs static analysis on skill directories to verify structure and frontmatter compliance. It does not execute the contents of the files it validates.
[DATA_EXPOSURE]: The skill documents security considerations in references/security-and-governance.md, specifically instructing authors to avoid secret scraping and to use restricted tool permissions (allowed-tools) to minimize the impact of potential misinterpretations.
[PROMPT_INJECTION]: The skill uses conditional and qualified language appropriately to preserve agent judgment across different usage contexts, as detailed in references/assumptions-and-adaptability.md. It provides templates to help users avoid common errors and hallucinations.

write-skill