hugging-face

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

• Third-party content exposure detected (high risk: 0.75). Outsider source: the Hugging Face actions (e.g., “get daily papers”, dataset/model reads) can fetch and ingest public/community-authored text from Hugging Face into the agent’s LLM context at runtime via the ClawLink CLI.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs running runtime npx commands that fetch and execute the remote ClawLink CLI (npx @useclawlink/cli, e.g. https://www.npmjs.com/package/@useclawlink/cli), which is a required dependency and would run remote code that controls agent actions/prompts.