YaOS / Platform / Developer Experience / Coding Harness / Team Operating Model
Operating Model
Team Operating Model
Clear ownership across humans, coding agents and deterministic automation.
Responsibility split
| Humans own | Agents own | Automation owns |
|---|---|---|
| Requirements Product interpretation Architecture judgment Risk acceptance Final approval | Codebase investigation Implementation Tests Mechanical refactors Verification evidence First-pass review Docs updates | Formatting Linting Typing Architecture constraints Security checks CI execution Harness validation |
Operating principle: Human = intent + judgment. Agent = execution + reasoning. Automation = enforcement.
Standard lifecycle
- Task arrives with goal, acceptance criteria and material constraints.
- Agent orients itself in the repo and reads only relevant context.
- Task is classified S/M/L/X.
- Agent implements in a branch/worktree.
- Verification progresses narrow → broad.
- Material changes receive an independent review context.
- PR contains evidence, assumptions and risks.
- CI enforces non-negotiables.
- Repeated review feedback is promoted into lint, rule, skill or documentation.