implement-with-subagents
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions and configuration do not contain any malicious patterns such as prompt injection, data exfiltration, or obfuscation. The orchestration logic is designed with defensive procedures that prioritize controller oversight and result verification.
- [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests untrusted data from supplied tickets and plan tasks (Ingestion points: SKILL.md). It possesses capabilities for spawning subagents and modifying the repository through the required 'implement' skill (Capability inventory: Step 5, Step 6). However, the risk is effectively mitigated by the mandatory Implementation Procedure, which requires the controller to independently verify every commit range, diff, and verification evidence produced by the subagents before advancing (Sanitization: Step 8).
- [COMMAND_EXECUTION]: The skill manages implementation through subagents but enforces strict ownership boundaries. It mandates that remote mutations remain with the controller and that all subagent work be captured in task-scoped commits that are reviewed by the controller.
Audit Metadata