subagent-driven-development-conventions
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions require the agent to ingest external data from project memory files and codebase indices before delegating tasks to subagents. This creates a surface where malicious instructions embedded in the project files could be passed to and executed by subagents.
- Ingestion points: Durable context, memory files, and codebase-domain-map (SKILL.md).
- Boundary markers: None mentioned for the context packets passed to subagents.
- Capability inventory: The skill manages implementer subagents who perform implementation tasks (modifying code).
- Sanitization: No validation or sanitization of the project-specific context is specified.
Audit Metadata