subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill coordinates subagents to perform implementation, testing, and multi-stage code reviews. It does not contain any malicious patterns, hardcoded credentials, or unauthorized network operations.
- [INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface by ingesting external data (implementation plans) and subagent reports. However, it implements security best practices by instructing the 'Spec Compliance Reviewer' to independently verify code changes against requirements and explicitly warns against trusting the implementer's potentially inaccurate reports.
- Ingestion points: Implementation plan files and implementation reports processed in
SKILL.mdandspec-reviewer-prompt.md. - Boundary markers: Uses markdown headers to separate requirements from reports.
- Capability inventory: File writing, testing, and git commits (performed by delegated subagents).
- Sanitization: Relies on logical verification (code review) as a primary defense-in-depth measure.
Audit Metadata