subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill manages a workflow where an orchestrator agent ingests and processes data generated by subagents (implementers and reviewers).
- Ingestion points: The orchestrator agent processes implementation code and review results across several phases defined in SKILL.md.
- Boundary markers: The skill uses highly structured prompt templates (implementer-prompt.md, spec-reviewer-prompt.md, code-quality-reviewer-prompt.md) to define clear Markdown sections for agent communication.
- Capability inventory: The skill utilizes the Agent tool for task delegation and describes the execution of project-specific verification commands like linting, testing, and building in Phase 6.
- Sanitization: The risk of malicious output is mitigated by a mandatory two-stage review gate (Phases 3 and 4) that forces independent validation of all work before it is accepted or Integrated.
- [SAFE]: The skill design is security-positive, as Phase 4 explicitly tasks a reviewer agent with auditing code for security vulnerabilities, including injection flaws, sensitive data exposure, and improper input validation. The skill further enforces safety through iterative fix cycles and strict adherence to a task-independence decision table to prevent race conditions or integration errors.
Audit Metadata