auto-decorrelation

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

• Third-party content exposure detected (medium risk: 0.30). The skill workflow recruits external cross-family verifier sidecars (like codex and agy) and sends them payloads containing code diffs and markers, which represents ingestion/handling of potential outsider-authored free text via sidecar review findings.