wiki-ingest

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches arbitrary public URLs ("User shares a URL and asks to capture it into the wiki" and Step 2: "URL: Fetch via WebFetch, then write a local copy"), and that fetched, untrusted content is read and interpreted (Step 2→Step 3 synthesis) and then directly drives actions (page composition, index updates, backlink audits and atomic applies in Steps 4–6), so third‑party pages could inject instructions that materially influence the agent's subsequent tool use and writes.