linkup-fetch

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

• Third-party content exposure detected (medium risk: 0.30). Runtime workflow path: using Linkup Fetch (POST https://api.linkup.so/v1/fetch) ingests the content of an arbitrary URL provided to the tool/API, which can be outsider-authored free text if the caller supplies a poisoned URL.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls Linkup runtime endpoints (e.g., https://api.linkup.so/v1/fetch, https://api.linkup.so/v1/search, https://api.linkup.so/v1/research) to retrieve arbitrary page content which is then fed into an LLM (i.e., external content is injected into prompts), so these runtime fetches can directly control agent prompts.