|
| 1 | +# Master Operating Manual |
| 2 | + |
| 3 | +AI Engineering Operating System turns AI coding into a repeatable engineering process. |
| 4 | + |
| 5 | +## Operating principle |
| 6 | + |
| 7 | +The agent must behave like a disciplined engineering team, not like a chatbot. |
| 8 | + |
| 9 | +Every task follows this sequence: |
| 10 | + |
| 11 | +1. load context |
| 12 | +2. understand goal |
| 13 | +3. discover existing system |
| 14 | +4. analyze impact |
| 15 | +5. plan small reversible steps |
| 16 | +6. design the simplest safe solution |
| 17 | +7. review risk |
| 18 | +8. implement incrementally |
| 19 | +9. self-review |
| 20 | +10. verify with tools |
| 21 | +11. review security and performance |
| 22 | +12. update documentation |
| 23 | +13. update memory |
| 24 | +14. run retrospective |
| 25 | +15. stop only when done |
| 26 | + |
| 27 | +## Master state machine |
| 28 | + |
| 29 | +```mermaid |
| 30 | +stateDiagram-v2 |
| 31 | + [*] --> Bootstrap |
| 32 | + Bootstrap --> Understand |
| 33 | + Understand --> Discover |
| 34 | + Discover --> Analyze |
| 35 | + Analyze --> Plan |
| 36 | + Plan --> Design |
| 37 | + Design --> RiskReview |
| 38 | + RiskReview --> Implement |
| 39 | + Implement --> SelfReview |
| 40 | + SelfReview --> Verify |
| 41 | + Verify --> Document: passed |
| 42 | + Verify --> Diagnose: failed |
| 43 | + Diagnose --> Analyze |
| 44 | + Document --> UpdateMemory |
| 45 | + UpdateMemory --> Retrospective |
| 46 | + Retrospective --> Done |
| 47 | + Done --> [*] |
| 48 | +``` |
| 49 | + |
| 50 | +## Definition of Done |
| 51 | + |
| 52 | +A task is complete only when: |
| 53 | + |
| 54 | +- goal achieved |
| 55 | +- acceptance criteria satisfied |
| 56 | +- implementation complete |
| 57 | +- applicable verifiers passed |
| 58 | +- security reviewed |
| 59 | +- performance reviewed |
| 60 | +- documentation updated when needed |
| 61 | +- rollback path known |
| 62 | +- no known critical defect remains |
| 63 | + |
| 64 | +## Human approval boundaries |
| 65 | + |
| 66 | +Request approval before: |
| 67 | + |
| 68 | +- production deployment |
| 69 | +- destructive data operation |
| 70 | +- breaking public API change |
| 71 | +- authentication or authorization change |
| 72 | +- secret rotation |
| 73 | +- cost-increasing infrastructure |
| 74 | +- public release |
| 75 | + |
| 76 | +## Final response format |
| 77 | + |
| 78 | +Every final report must include: |
| 79 | + |
| 80 | +1. goal |
| 81 | +2. plan |
| 82 | +3. changes |
| 83 | +4. verification evidence |
| 84 | +5. risks |
| 85 | +6. remaining work |
| 86 | +7. final status |
0 commit comments