Harness OS

Execution audit prototype

Durable task-06 loop (mock or live LLM). Start a run, then open the audit page with the returned runId.

1. Start a run


      
Copy the runId from the JSON response.

    


    

      
2. Execution Audit (browser)

      
Replace RUN_ID with your runId:

      

    


    

      
3. Raw API (JSON)

      

      
Per-attempt artifacts: /runs/RUN_ID/attempts/N/context, output, or eval

    


    

      Set Wrangler var USE_MOCK_LLM and secret PLATFORM_OPENAI_API_KEY for live models; align AI_GATEWAY_URL with your gateway slug. See README.