fanout

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

• Third-party content exposure detected (medium risk: 0.30). During run-everything (and main-thread fallback), the skill constructs a Workflow that has leaf/slice agents return free-text findings from the diff, then a Stage-0 “extraction” agent ingests those raw outsider-authored review texts (i.e., agent-generated free text from repository-diff context) via the extractInput string under the ### Surface: headers.