skill-optimizer
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process the content of third-party SKILL.md files and related documentation. This creates an inherent attack surface where malicious instructions contained within the analyzed skills could attempt to subvert the optimizer's logic. \n
- Ingestion points: The skill reads the target SKILL.md and referenced files as part of its 'Review and Select Materials' workflow step (SKILL.md). \n
- Boundary markers: There are no technical delimiters (e.g., XML tags or specific 'ignore' blocks) specified in the instructions for the content of the ingested files, although the skill includes a high-level rule to only review and not execute the business tasks of the target. \n
- Capability inventory: The skill possesses the capability to read and write to the local filesystem and execute verification scripts (SKILL.md, references/review-checklist.md). \n
- Sanitization: While the skill guidelines explicitly prohibit echoing sensitive data found during analysis, there is no technical sanitization of the input text to prevent the model from obeying instructions potentially embedded in the markdown of optimized skills.
Audit Metadata