image-compare
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the execution of a custom Python script (imgdiff.py) to analyze images. This process is local and does not involve network communication.\n- [EXTERNAL_DOWNLOADS]: The instructions involve installing well-known and trusted Python packages (pillow, numpy, scikit-image) and optionally a Node.js package (pixelmatch), all from official registries.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided images, which serves as a potential vector for malicious content intended to influence the agent.\n
- Ingestion points: User-provided image files (before.png, after.png) processed in imgdiff.py and subsequently viewed by the agent.\n
- Boundary markers: The skill does not implement specific boundary markers or warnings to the agent regarding the untrusted nature of the image content.\n
- Capability inventory: The skill is limited to local file system reads and writes (diff image generation) without network access.\n
- Sanitization: Images are processed mathematically, but the visual content is not sanitized against adversarial text or patterns before agent review.
Audit Metadata