skill-review
Pass
Audited by Gen Agent Trust Hub on Feb 25, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements structural validation and metadata analysis using local scripts. Review of scripts/validate.py and scripts/analyze.sh confirms they perform legitimate file parsing and do not contain malicious code, network requests, or unauthorized system access.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it is designed to ingest and interpret content from user-specified skill directories. 1. Ingestion points: SKILL.md and script files within the target directory. 2. Boundary markers: The skill reads file content directly without specific markers or instructions to ignore embedded prompts. 3. Capability inventory: Permission to read files and interactively apply suggested improvements (file writing). 4. Sanitization: No explicit sanitization or filtering of the ingested skill content is performed before analysis.
Audit Metadata