usv240/argus-detection-evolution ? reverse-engineered prompt

Build me ARGUS, a web app that shows an attacker AI and a defender AI battling each other over real Splunk data, so I can see where my detections are weak and how they improve over time.

I want a clean dashboard where I can pick a security scenario, start a run, and watch the red team invent evasive attack variants, the evaluator score them live against Splunk, and the blue team rewrite the detection rule to catch more of them without creating false positives. Show the baseline and evolved SPL, the coverage gain, what still slips through, and a simple proof panel with the live numbers behind the result.

It should also let me review the proposed detection, approve or reject it, and export the final rule as a Splunk app bundle with a certificate and validation report. Keep the UI polished and easy to follow, and use real live queries and computed results, not mocked data. If you need current Splunk or Anthropic docs, look them up online first.