5-steps
Pass
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is purely instructional and contains no code, shell commands, or network operations.
- [PROMPT_INJECTION]: The skill instructions involve analyzing user-provided requirements and existing codebases in SKILL.md (Step 2), which act as untrusted ingestion points. No explicit boundary markers are defined within the prompt. However, the skill does not possess dangerous tools or autonomous execution capabilities (capability inventory) and instead mandates human approval for deletions or design changes, maintaining a safe posture.
Audit Metadata