subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Jul 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: Extensive analysis of the orchestration logic, file handling, and subagent prompt templates revealed no malicious behavior, unauthorized data access, or persistence mechanisms.
- [PROMPT_INJECTION]: The skill handles untrusted data from external implementation plans and subagent-generated reports, creating an indirect prompt injection surface. This risk is addressed by the architectural design, which isolates subtasks within separate agent sessions and utilizes structured prompt templates that clearly delineate instructions from data (e.g., using file paths rather than inline text concatenation).
- [COMMAND_EXECUTION]: The skill utilizes standard version control systems (Git and Jujutsu) and file system operations to manage code implementation and review. These actions are strictly within the scope of the skill's declared purpose and do not involve unauthorized privilege escalation or suspicious remote execution.
Audit Metadata