product-design
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill logic enforces a controlled implementation workflow where the agent must first ground itself in the product context, provide a critique, and propose a plan that requires user approval before any file modifications occur (Workflow Phase 4 and 5). This prevents automated or unauthorized changes.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it is instructed to ingest and process potentially untrusted external data such as product documentation, PRDs, and issue descriptions.
- Ingestion points: Workflow Phase 1 in SKILL.md (docs, PRDs, issues, README, and browser state).
- Boundary markers: Absent. No explicit delimiters or instructions are provided to help the agent distinguish between its instructions and the content of the analyzed documents.
- Capability inventory: The skill has the ability to read local files for inspection and write to the filesystem during the implementation phase.
- Sanitization: Absent. There is no instruction to filter, validate, or sanitize the content of the processed files before they are incorporated into the agent's context.
Audit Metadata