subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection by processing external implementation plans and requirements. * Ingestion points: Task descriptions are extracted from plan files and interpolated into subagent prompts in implementer-prompt.md and spec-reviewer-prompt.md. * Boundary markers: The task content is delimited using Markdown headers (e.g., ## Task Description), which provide limited isolation against adversarial instructions. * Capability inventory: The implementer subagents have the capability to modify the codebase, write tests, and commit changes. * Sanitization: No sanitization of plan content is performed.
Audit Metadata