openrouter-images
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Documentation in SKILL.md instructs the user or agent to execute 'npm install' to set up dependencies and 'npx tsx' to run the TypeScript scripts.
- [EXTERNAL_DOWNLOADS]: The skill downloads the 'tsx' package from the NPM registry and makes network requests to 'https://openrouter.ai/api/v1/chat/completions' to interact with the image generation models.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8).
- Ingestion points: Untrusted data enters the agent context via the text prompt argument in 'generate.ts' and 'edit.ts', and through local image files read in 'edit.ts'.
- Boundary markers: The scripts do not implement delimiters or specific 'ignore' instructions when interpolating user-provided text or image data into the API request body.
- Capability inventory: The 'lib.ts' utility includes 'readFileSync' for file access, 'writeFileSync' for saving images to the local system, and 'fetch' for external network communication.
- Sanitization: There is no evidence of validation, escaping, or filtering of the prompt or image content prior to its transmission to the model.
Audit Metadata