legal-policy-drafter
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes the entire repository and external documentation to map product behavior and data flows. This creates a surface where malicious instructions embedded in the repository (e.g., in READMEs or code comments) could potentially influence the drafted policies.
- Ingestion points: All repository files (SKILL.md step 1-4) and external documentation retrieved via web search.
- Boundary markers: The instructions do not specify explicit separators for ingested content, but emphasize evidence-based drafting and the use of 'TODO(confirm)' markers for validation.
- Capability inventory: File reading, web searching, and file writing to the repository root.
- Sanitization: The skill relies on human review and explicit confirmation markers rather than automated sanitization of ingested content.
- [EXTERNAL_DOWNLOADS]: The skill researches third-party providers by fetching current official documentation, DPAs, and subprocessor lists from the web to ensure privacy disclosures are accurate and up-to-date.
Audit Metadata