Workflows

YarraMate supports two entry points and one ongoing maintenance loop.

Bring an existing system into the model

There is no CLI verb that discovers a system on its own. The coding-agent skill inspects repository evidence, proposes native documents, and leaves the decision with you:

yarramate check .yarramate/workspace.yaml --json
yarramate reconcile .yarramate/workspace.yaml

Review every proposed claim before treating it as canonical. See integrations for installing the skill.

Design before implementation

yarramate design .yarramate/workspace.yaml

The interview works from goals and constraints toward concepts, responsibilities, relationships, and states. Resume it later with the same command; repository documents carry context between sessions. Add --facilitate to render the same interview in plain language for a stakeholder workshop away from raw YAML.

Maintain the model with the code

yarramate ask .yarramate/workspace.yaml
yarramate check .yarramate/workspace.yaml --json
yarramate reconcile .yarramate/workspace.yaml

• ask orients a person or agent before a change.
• check validates native documents deterministically.
• reconcile compares modeled claims with attached evidence.

When a design changes, edit the native YAML alongside the implementation. Regenerate projections; never repair generated output by hand.

Keep evidence honest

Evidence can confirm, challenge, or leave a claim unknown. Unknown is useful: it identifies where stronger proof is needed without pretending the implementation was inspected.

Next: CLI reference →

These adoption guides are maintained by the website team. Read the normative product documentation →