facebook

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow uses ClawLink to operate Facebook via browser/OAuth and then fetches Facebook Page/post/comment text (outsider-authored free-form content) which is ingested into the agent’s LLM context through the CLI’s “describe/get/list” outputs.

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

• Potentially malicious external URL detected (high risk: 0.90). Yes — the skill instructs running the remote npm CLI at runtime via "npx @useclawlink/cli" (see https://www.npmjs.com/package/@useclawlink/cli), which fetches and executes external code that then performs/controls Facebook actions on behalf of the agent.