stage2_plan.md

Stage 2: Data Aggregation & Decision Logic

Goal

Implement the core intelligence of Agent Zero: aggregating data from various sources (CodeRabbit, Vercel, Tests) and calculating the Merge Confidence Score (M.C.S.).

1. M.C.S. Calculation Script (scripts/calculate_mcs.py)

• Inputs: data.json containing:
  • coderabbit: status, summary
  • vercel: deployment status, url
  • codecov: coverage %
  • shadow_agent: analysis result
• Logic:
  • Base Score: 0
  • +40 for CodeRabbit Approval
  • +30 for Vercel Success
  • +20 for High Test Coverage (>80%)
  • +10 for Clean Shadow Agent Audit
  • -20 for Critical Build Failed
• Output: JSON with { "mcs": int, "status": "MERGE_CANDIDATE" | "AUTOCORRECT" | "NEEDS_REVIEW" }

2. Kestra Integration

• New Flow Tasks:
  • fetch_vercel_status: HTTP Request to Vercel API (Mocked for now)
  • fetch_code_quality: Parsing CodeRabbit/Codecov inputs (Mocked)
  • run_mcs_calculation: Python script task running calculate_mcs.py
  • decision_switch: Switch case based on MCS status.

3. Mock Data for Testing

• Create tests/mock_data/happy_path.json (MCS 100)
• Create tests/mock_data/broken_build.json (MCS 10)