This repository operates as a control loop for autonomous agent development.

• pass_at_1 target: 100% (all turbo tasks pass)
• merge_cycle_time target: < 30 min
• revert_rate target: < 5%
• human_intervention_rate target: < 10%

• CI checks (GitHub Actions)
• turbo build / turbo check-types / turbo lint outcomes
• Web E2E outcomes (Playwright)
• Biome static analysis
• Control audit results

• Gate sequence: smoke → check → test
• Retry budget: 2
• Escalation: human_oncall after retry budget exhausted

• Code edits
• Package configuration updates
• Script updates
• Policy updates
• Documentation updates
• CI workflow updates

1. Measure: Run sensors (build, lint, test, audit)
2. Compare: Check against setpoints
3. Decide: Identify failing gates and root causes
4. Act: Apply fixes via actuators
5. Verify: Re-run sensors to confirm resolution