subagent-driven-development
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it processes task descriptions and plans. 1. Ingestion points: Plan files (SKILL.md) and task text (implementer-prompt.md). 2. Boundary markers: Markdown headers are used to delimit external data. 3. Capability inventory: Subagents have the ability to write code, execute tests, and perform git operations (implementer-prompt.md). 4. Sanitization: The workflow relies on independent reviewer subagents (spec-reviewer-prompt.md, code-quality-reviewer-prompt.md) to catch discrepancies.
- [NO_CODE]: The skill is purely informational, consisting of markdown files and prompt templates without any executable scripts or binaries.
Audit Metadata