The Agent Skills Directory

SAFE (SAFE): The repository consists of templates and instructional material designed to standardize skill development. No malicious prompt injection or obfuscated code was found.\n- SECURITY_GOVERNANCE (SAFE): The inclusion of references/security-and-governance.md demonstrates a security-first approach, providing checklists for script safety, tool restrictions, and human-in-the-loop requirements for high-stakes actions.\n- LOCAL_VALIDATION (SAFE): The scripts/validate_skill_dir.ts utility is a diagnostic tool for linting skill metadata and structure. It uses standard file system APIs and performs no network operations or data exfiltration.\n- INDIRECT_PROMPT_INJECTION (SAFE): While templates like SKILL.fork-task.template.md define surfaces for argument ingestion, they are for developer use in building specific skills. The templates themselves do not provide an active attack surface or bypass safety guardrails.

write-skill