humanloop

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

• Third-party content exposure detected (high risk: 0.75). The required workflow uses the ClawLink CLI to connect to Humanloop and then run/list actions, which at runtime will fetch and ingest Humanloop project/experiment/session data (outsider-authored free text from the Humanloop workspace) into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs running npx @useclawlink/cli (https://www.npmjs.com/package/@useclawlink/cli), which fetches and executes remote npm package code at runtime to control agent actions (login/connect/actions/run), so it is a required runtime dependency that executes remote code.