review-loop
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute local npm scripts (
draft,sheet,still,build,wave) to manage rendering and analysis of visual project files. - [INDIRECT_PROMPT_INJECTION]: The agent performs a visual audit on assets such as
out/sheets/beats-N.pngandout/still.png, which are considered external data sources for the model's critique. - Ingestion points: Visual assets generated in the project's
out/directory. - Boundary markers: No delimiters or safety warnings are used when the agent is directed to open and inspect the generated files.
- Capability inventory: The skill body is limited to running local rendering scripts; it does not request network access or sensitive file operations.
- Sanitization: No sanitization is performed on the visual assets prior to the agent's analysis.
Audit Metadata