unirig

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). Outsider-authored free text can only reach the LLM indirectly if it is provided via the user-controlled --input/--input-dir assets (e.g., model files) and then ingested by UniRig’s upstream inference code; the provided runtime workflow itself does not read outsider text sources like emails/chat/issue bodies.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches and executes required upstream code and model weights at runtime (cloning/using https://github.com/VAST-AI-Research/UniRig and auto-downloading the checkpoint from https://huggingface.co/VAST-AI/UniRig), and the installer also registers the skill via https://github.com/akillness/jeo-skills — all runtime fetches of remote code/weights that the pipeline depends on.