subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious behavior, obfuscation, or data exfiltration detected. The skill defines a structured workflow for development using internal agent capabilities.
- [PROMPT_INJECTION]: The skill is designed to ingest and execute tasks from external implementation plans, which creates a surface for indirect prompt injection. This risk is mitigated by the architectural design which includes independent specification and quality review subagents to verify all generated code before completion. Ingestion points: Task descriptions are read from implementation plan files and interpolated into subagent prompts. Boundary markers: Prompt templates do not use explicit delimiters, but provide clear section headers. Capability inventory: The implementer subagent is granted file-write and test execution permissions. Sanitization: The workflow enforces a two-stage review process (spec compliance and code quality) as a mandatory gate for all implemented code.
Audit Metadata