subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Apr 24, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, unauthorized command execution, or data exfiltration attempts were found. The skill incorporates quality and security controls by enforcing fresh subagent contexts for each task and requiring a two-stage review process (specification compliance and code quality) before task completion.
- [PROMPT_INJECTION]: The skill design creates a surface for indirect prompt injection by interpolating external task descriptions from plan files directly into subagent prompt templates. However, this is consistent with the primary functional purpose of the tool and does not indicate malicious intent.
- Ingestion points: Plan files read by the controller (e.g., docs/superpowers/plans/feature-plan.md).
- Boundary markers: Task text is inserted into prompts without explicit delimiters or protective instructions for the subagent.
- Capability inventory: Subagents are instructed to modify codebase files, execute tests, and commit changes.
- Sanitization: The task content is not sanitized or escaped before interpolation.
Audit Metadata