Autonomous Agent Team

This directory defines the specialized agent team for 24/7 MUTX shipping.

Core Operating Model

• One orchestrator plans and dispatches work.

• Specialists own bounded file areas and do not overlap without an explicit handoff.

• Every code-writing agent opens a branch and pull request instead of pushing to main.

• Every PR requires a second agent reviewer plus CI before merge.

• Low-risk lanes can auto-merge after review; risky lanes stop at staging or require human approval.

Team Members

• mission-control-orchestrator

• control-plane-steward

• operator-surface-builder

• auth-identity-guardian

• cli-sdk-contract-keeper

• runtime-protocol-engineer

• qa-reliability-engineer

• infra-delivery-operator

• observability-sre

• docs-drift-curator

Launch Order

1. mission-control-orchestrator

2. qa-reliability-engineer

3. cli-sdk-contract-keeper

4. control-plane-steward

5. operator-surface-builder

6. auth-identity-guardian

7. observability-sre

8. infra-delivery-operator

9. runtime-protocol-engineer

10. docs-drift-curator

Shared Rules

• Trust code over docs when they disagree.

• Prefer the smallest correct change.

• Treat route drift as a bug.

• Do not use npm run lint as a required gate until the ESLint setup is repaired.

• Use npm run build, Python checks, targeted pytest, and targeted Playwright or infra validation as the real gates.

• If API contracts change, update dependent surfaces in the same workstream or open linked tasks immediately.