patent-review
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (patent drafts) which could contain malicious instructions designed to subvert agent logic.
- Ingestion points: Phase 1 in
SKILL.mdingests patent content via direct text paste or extraction from a.docxfile. - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the input as untrusted data or to ignore any instructions embedded within it.
- Capability inventory: The skill has the capability to write reports to the local filesystem (Phase 4 in
SKILL.md) and perform external network searches for literature (Phase 5 inSKILL.md). - Sanitization: The skill does not implement validation or sanitization of the patent text before using it in the interactive review process or potentially interpolating it into search queries or file names.
Audit Metadata