Alert Evidence Review

Publisher: HOK MING SHUM ยท Community-maintained Alert Lab project

Review sanitized HTTP-400/rejection evidence under a disclosed, illustrative reference policy. Replay recommendations and compare why they changed. This is not general incident diagnosis, live monitoring, production paging guidance, or automatic remediation.

Connect the anonymous remote MCP endpoint at /mcp. Tools: replay_evidence and compare_replays. Read alert-lab://policy/current for the schema and thresholds.

Supply normalized numeric/categorical evidence and anonymous case labels only. Missing signals stay unknown. Classifier values and deadlines are simulated. No model or OpenAI API key is needed by this service.