Port uses evaluation in the broad product sense: command proofs, board evidence, and lane-specific validation that demonstrate the product still behaves as claimed.
For repository-level confidence, run:
just missionThat covers:
- formatting
- workspace tests
- doctests
- board health
- current mission signal
Every story should leave behind:
- automated tests where practical
- command proofs for user-facing surfaces
- board-linked evidence logs under the story directory
- documentation changes when the operator contract changes
| Area | Primary signal |
|---|---|
| Repo health | just mission, just keel doctor |
| Local Firecracker | port doctor, local artifact build, `machine launch |
| Hosted standard lane | control-plane plus node-agent round trip |
| Cloud Hypervisor | lane-specific local or hosted round trip |
| AVF | AVF launcher plus canonical machine and guest verbs |
| Firecracker/PVM | prepared-node contract and explicit host-kit validation |
| Services | `port service secret |
| Docs and help | command proofs and targeted doc audits |
Top-level docs are part of the product surface.
README.mdshould stay short and point to canonical detail.CONFIGURATION.mdshould hold detailed config and workflow examples.port --helpshould remain concise and useful at a glance.
Add or update proofs when a change affects:
- the operator-visible command contract
- config shape or environment-variable behavior
- runtime ownership or hosted routing
- platform support boundaries
- artifact or service lifecycle semantics