verify

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

• Third-party content exposure detected (medium risk: 0.30). In workflows/verify.mjs the Matrix/Spec/Defect-guessing/Judging/Fix agents read and ingest user/outsider-authored free text from the repo itself at runtime (e.g., planText/planPath and diffCmd output from the caller-controlled working tree), then feed it into agent(...) prompts that drive decisions—creating an indirect prompt-injection path via plan/spec or code+comments.