subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as an orchestrator that takes external feature descriptions and context to decompose them into tasks for specialized sub-agents, creating a vulnerability surface.
- Ingestion points: Feature names, subtask descriptions, and background context provided by users are ingested in Step 1 (Decompose) and Step 4 (Delegation Protocol).
- Boundary markers: The templates provided for task delegation do not include explicit delimiters or instructions for sub-agents to ignore potentially malicious or contradictory commands embedded within the feature context.
- Capability inventory: The agents orchestrated by this skill are expected to perform file writes (code and migrations), database schema modifications, and security policy configurations (RLS).
- Sanitization: There are no instructions for sanitizing or validating external input before it is interpolated into the prompts for sub-agents.
Audit Metadata