AI-OS governance is based on explicit goals, human approval gates, verification evidence, and reversible change.
flowchart TD
Proposal[Proposal] --> Impact[Impact Analysis]
Impact --> Risk[Risk Classification]
Risk --> Plan[Implementation Plan]
Plan --> Approval{Human Approval Required?}
Approval -- Yes --> Human[Maintainer Approval]
Approval -- No --> Implement[Implement]
Human --> Implement
Implement --> Verify[Verify]
Verify --> Pass{Pass?}
Pass -- No --> Rework[Rework]
Rework --> Verify
Pass -- Yes --> Record[Record Decision]
Record --> Done[Done]
- Release process changes
- Security policy changes
- Governance changes
- Breaking changes to master prompt semantics
- Automation that can mutate repositories
- Anything that increases risk or cost
Use ADRs for structural decisions. ADRs must explain context, decision, alternatives, consequences, and rollback.