subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Feb 18, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- Prompt Injection (LOW): Indirect Prompt Injection Surface.
- Ingestion points:
implementer-prompt.mdandspec-reviewer-prompt.mdingest untrusted task descriptions and reports. - Boundary markers: The prompts use Markdown headers as delimiters but lack explicit instructions to disregard potential commands or safety overrides embedded in the task content.
- Capability inventory: The process empowers subagents to perform file writes (implementation), execute tests (subprocess), and commit code (git operations).
- Sanitization: No evidence of sanitization or escaping for the interpolated plan text is provided in the templates.
Audit Metadata