prototype
Pass
Audited by Gen Agent Trust Hub on Nov 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided ideas and selection strings to manage file system operations, creating a surface for indirect prompt injection.
- Ingestion points: User input defining the 'idea' (Step 1) and the 'pick line' string copied from the browser and pasted back to the agent (Step 5).
- Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are defined for the user-provided input strings.
- Capability inventory: The skill utilizes file-writing to create component variants and file-deletion to 'prune' losing variants, switcher scripts, and route files.
- Sanitization: There is no evidence of sanitization or validation of the 'pick line' or user-provided labels before they are used to determine which files to delete.
Audit Metadata