prior-art-search
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the Bash tool to run Python search utilities that interface with BigQuery. This is the primary function of the skill and utilizes official libraries.
- [DYNAMIC_EXECUTION]: The prior_art_types.py script performs dynamic path manipulation by inserting the project root into sys.path to resolve internal module imports.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external patent data from BigQuery, creating an attack surface where text within a patent could theoretically target the agent during report generation.
- Ingestion points: Patent data is retrieved in steps 3 and 5 of the methodology defined in SKILL.md.
- Boundary markers: None explicitly defined in the methodology for isolating or quoting patent content to prevent instruction following.
- Capability inventory: The agent has access to Bash, Read, and Write tools.
- Sanitization: The methodology relies on standard LLM generation without specialized escaping for retrieved patent text.
Audit Metadata