Tutorial: First Topic Bootstrap
This tutorial walks you through bootstrapping Nimi Coding into a new project and getting .nimi/** set up. By the end you will have:
@nimiplatform/nimi-codinginstalled.nimi/**populated with package-owned bootstrap source- A managed
AGENTS.md/CLAUDE.mdblock confirming the bootstrap - An authoritative JSON handoff payload for
spec_reconstruction
Prerequisites
| Requirement | Why |
|---|---|
| Node.js 24 or newer | Package runs on Node |
| Project root with version control | Bootstrap will create new files; you'll want to review them |
| An admitted external AI host | Required for the next step (spec_reconstruction) |
The tutorial does not require the spec to be reconstructed yet — that's the next step after bootstrap.
Step 1: Install The Package
Install @nimiplatform/nimi-coding as a dev dependency:
npm install --save-dev @nimiplatform/nimi-codingor:
pnpm add -D @nimiplatform/nimi-codingThe command can vary by package manager, but the result should be the same: nimicoding is available as a project CLI.
After installation, npx nimicoding --help should print help text.
Step 2: Run nimicoding start
nimicoding start is the bootstrap entry. It is interactive: it explains each step, asks for confirmation, applies the step.
The CLI walks through:
- Detect project state.
- Confirm or accept managed AI entrypoints (
AGENTS.md,CLAUDE.mdblocks). - Project package source into project paths.
- Seed
.nimi/spec/_metaand bootstrap files. - Update
.gitignorefor local runtime state. - Prepare a JSON handoff payload for
spec_reconstruction. - Print a paste-ready prompt directly in the terminal.
After this step, the project root has:
| Path | Contents |
|---|---|
.nimi/methodology/ | Methodology source (policies) |
.nimi/contracts/ | Schema source |
.nimi/config/ | Bootstrap config |
.nimi/spec/_meta/ | Spec generation metadata seed |
AGENTS.md (or block within) | Managed Nimi Coding block |
CLAUDE.md (or block within) | Managed Nimi Coding block |
.gitignore | Updated with local-state ignore patterns |
Step 3: Verify With nimicoding doctor
nimicoding doctor validates that the bootstrap is in a healthy state.
It checks:
.nimi/**bootstrap seed presence.nimi/local/and.nimi/cache/exist and remain ignored- Bootstrap contract compatibility metadata
- Cross-contract reference alignment
- Host-adapter boundary truth
- Skill result-contract alignment
- Handoff context-order readiness
A healthy doctor output confirms you can hand off skills to an external AI host.
If something is yellow / red, the report names the area; address the named issue and re-run.
Step 4: Examine The Generated Handoff Payload
nimicoding start produced a JSON handoff payload for spec_reconstruction. The file lives under .nimi/local/handoff/ (or wherever the CLI placed it; the CLI output names the path).
Open the JSON. Notice:
skill: "spec_reconstruction"- A typed payload with required context order
- Source basis paths the host should read
The payload is the contract between the package and the host.
Step 5: Stop Here
You have completed the bootstrap. The project is now Nimi Coding-adopting. The next step (in First Wave End-To-End) is to actually run a wave.
For this tutorial, stop here. Do not yet hand the payload to a host; do not yet run reconstruction. Bootstrap is complete; that is your outcome.
Common Issues
| Symptom | Resolution |
|---|---|
nimicoding start complains about existing files | Review the conflict; the CLI preserves project-owned truth and refuses to overwrite |
nimicoding doctor reports lifecycle drift | Run start again; the CLI is idempotent for bootstrap |
| AGENTS.md changes not applied | The CLI shows confirmation; if you declined, re-run and confirm |
What You Have Now
After this tutorial, your project has:
- A complete
.nimi/**bootstrap. - Managed AI entrypoints in
AGENTS.mdandCLAUDE.md. - A
spec_reconstructionhandoff payload ready to hand off.
What you do not yet have:
- A reconstructed canonical spec tree (next step).
- Any topic / wave / packet artifacts (created later as needed).
- AI-coding work governed yet (you have the discipline available but no work in progress).
Next Step
If you want to actually use the methodology on a sample task, continue to First Wave End-To-End.
What This Tutorial Does Not Cover
This tutorial does not cover host-specific configuration for specific AI hosts. The package is host-agnostic; you can pick any admitted host.