design-driven-dev
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill follows a strict human-in-the-loop policy. It only proceeds with development tasks after the user has reviewed and approved specific documentation phases. It handles documentation directory discovery and initialization safely, requesting user confirmation before selecting or creating directories.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from local design documents to guide code generation and maintain alignment. 1. Ingestion points: HLD, LLD, and EARS documentation files located in the project's DOCS_DIR, as specified in SKILL.md. 2. Boundary markers: None present. 3. Capability inventory: The skill performs file system read/write operations and provides implementation suggestions across all phases defined in SKILL.md. 4. Sanitization: None present. This vulnerability surface is assessed as safe because the workflow incorporates frequent, mandatory user review checkpoints that mitigate the risk of the agent autonomously following malicious instructions embedded in processed data.
Audit Metadata