subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed for development workflow management and does not contain any executable code, remote downloads, or credential harvesting patterns. The prompt templates include robust safety instructions, such as advising reviewers to verify implementations independently rather than trusting subagent reports.
- [PROMPT_INJECTION]: The skill processes data from implementation plans to populate subagent prompts. This surface is a common design pattern for orchestration skills.
- Ingestion points: Implementation plan files are read in
SKILL.mdand their task descriptions are extracted for use in subagent prompts. - Boundary markers: The prompt templates do not use specific XML or unique delimiters for external task content, relying instead on standard markdown headers.
- Capability inventory: Subagents are dispatched using the
general-purposetool to perform implementation tasks, which typically includes file modifications and test execution. - Sanitization: The skill does not explicitly sanitize the extracted task text, but the multi-stage review process acts as a behavioral control.
Audit Metadata