breakdown-epic-arch
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external content which creates an indirect prompt injection surface.
- Ingestion points: Ingests external data through the
[The content of the Epic PRD markdown file]placeholder in SKILL.md. - Boundary markers: Absent. The external content is placed into the prompt without explicit delimiters or warnings to ignore embedded directives.
- Capability inventory: Includes file-writing capabilities as it requests the final document to be saved locally to
/docs/ways-of-work/plan/{epic-name}/arch.md. - Sanitization: Absent. There is no instruction to sanitize, validate, or filter the contents of the PRD before ingestion.
Audit Metadata