regulatory-prior-art-survey
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external regulatory documents and guidance from various web sources (e.g., EU Cellar, eCFR), which constitutes an attack surface for indirect prompt injection. 1. Ingestion points: External source URLs listed in the source registry and angle definitions. 2. Boundary markers: The skill instructs the agent to scan for and strip instructions like 'ignore previous instructions' and uses schema validation for all artifacts. 3. Capability inventory: The agent can execute local Python scripts via 'uv run', perform network requests for data retrieval, and write output files to the local file system. 4. Sanitization: Specific instructions mandate scanning for agent-directed notes and stripping content from 'AGENTS.md' files before reading.
- [COMMAND_EXECUTION]: The skill utilizes 'uv run' to execute local validation scripts (scripts/validate_regulatory_prior_art.py) to ensure the structural integrity and schema compliance of generated artifacts. This is a standard and well-contained operational pattern.
Audit Metadata