specification-writing
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external user-controlled data such as patent claims and invention disclosure documents, which introduces an indirect prompt injection attack surface.
- Ingestion Points: Ingests data from
patent/CLAIMS.md,patent/INVENTION_DISCLOSURE.md, andpatent/PRIOR_ART_REPORT.mdin the main workflow. - Boundary Markers: Absent; there are no specific encapsulation tags or instructions telling the agent to ignore instructions embedded within the ingested files.
- Capability Inventory: Possesses file system write capabilities (
Write,Edit,Bash) to create specification files underpatent/specification/and can invoke sub-agents or sub-skills. - Sanitization: Absent; the content is parsed and structured into sections without explicit validation, escaping, or sanitization rules.
Audit Metadata