patent-novelty-check
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process user-supplied invention descriptions and external reports, which are then passed to a high-reasoning sub-agent, creating a vector for instructions to be injected into the agent's logic.
- Ingestion points: Untrusted data enters the agent context through the
$ARGUMENTSvariable (invention description) and thepatent/PRIOR_ART_REPORT.mdfile (retrieved prior art). - Boundary markers: The skill lacks explicit boundary markers or instructions for the model to ignore any embedded commands or directives within the ingested invention details or research data.
- Capability inventory: The skill allows access to powerful tools including
Bash,WebSearch, andWebFetch, which significantly increases the impact if an injection attack is successful. - Sanitization: No sanitization, filtering, or validation is performed on the invention description or the contents of the prior art reports before they are interpolated into the prompt for the sub-agent analysis.
Audit Metadata