vault-wiki

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context during ingest: the workflow uses web_extract (or xurl for X) to fetch remote article text at runtime, then saves it into wiki/<target>/raw/... and proceeds to summarize/synthesize it into wiki pages—i.e., untrusted public content is ingested as readable prose.