and-implement

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

• Third-party content exposure detected (medium risk: 0.65). The required workflow reads the “claim record, ownership evidence, Package Contract, accepted and-clarify decisions, blockers, linked implementation artifacts” (step 1, plus “use the configured backend reference”), which are external/outsider-authored free-text fields originating from the delivery unit/trackers via the configured backend at runtime, creating an indirect prompt-injection channel into the agent’s LLM context.