human-eval-handoff-repair

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

• Third-party content exposure detected (high risk: 0.75). The required workflow ingests “Target public handoff package folder or ZIP” and “Current task schemas from the target UI data or target task CSVs,” which are outsider-authored public web/UI artifacts at runtime and may include free-text fields (e.g., evidence excerpts/notes) that the agent reads into the LLM context for QC and repair.