principle-build-the-lever
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions promote security best practices such as ensuring work is reviewable and deterministic, and requiring human-in-the-loop for external communication (e.g., drafting local text instead of sending automated messages).
- [PROMPT_INJECTION]: Includes specific instructions for the agent to follow regarding attribution and subagent constraints ("no agent may communicate with a person", "never add AI attribution"). These are functional guidelines for task execution and do not constitute malicious injection or safety filter bypasses.
- [DYNAMIC_EXECUTION]: The skill encourages the creation of tools, scripts, and codemods to perform work. These are intended to be local, deterministic artifacts for human review and verification rather than vectors for arbitrary code execution from untrusted sources.
- [SAFE]: References to the Cursor plugins repository in the third-party notices represent legitimate attribution for adapted material from a well-known service.
Audit Metadata