The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill follows best practices for agent instructions by using structured data, approval gates, and human-in-the-loop validation for its optimization tasks.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill only references its own internal asset and reference files for templates and checklists. No evidence of unauthorized file access, hardcoded credentials, or external data transmission was found.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to process external agent skills (untrusted data), it incorporates strong safety boundaries such as the 'Baseline Contract' and an 'Approval Gate' (Phase 4) which requires explicit human confirmation before any changes are applied, effectively mitigating the risk of being influenced by malicious content within the skills it analyzes.

skill-optimizer