dhanunjayavuppala/SIEMTriage — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a working SIEM triage app for SOC analysts. It should take a Microsoft Sentinel or Defender XDR incident ID, pull the alerts and entities, enrich things like users, devices, IPs, domains, and hashes, run useful hunting queries, then show a clear verdict with confidence, evidence, recommended actions, and a suggested deep dive plan.

I want a simple web UI with an incident queue, an incident detail page where I can watch the agent’s steps live, and a decision area where the analyst can choose what to do next. The agent must stay read only. It should never close incidents or take response actions by itself.

Please include a demo mode with sample incidents so I can run it without Azure, Postgres, or Redis first. Also include the full local stack with a database, queue, worker, ingest endpoint, and eval page that measures accuracy and false negatives against prior analyst decisions.

Use the current Claude Agent SDK and Microsoft docs if needed.

Want more depth? Deep Reverse