capability-evolver

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

• Third-party content exposure detected (high risk: 0.72). The required runtime workflow can ingest outsider-authored free text via EvoMap Hub–delivered mailbox messages (e.g., task_available, hub_event, asset_submit_result, skill_update) that the agent reads from the local Proxy mailbox and includes in its LLM context; these messages originate from third parties on the network, not the operating user.