Reproducible walkthrough

Run a repository change you can inspect

Use a disposable repository to add a health-check endpoint and tests while preserving the existing API. The workflow makes the plan, checks, recovery boundaries, and result visible.

The path from goal to evidence

Four deliberate steps

Prepare the repository, preview the change, and keep the result reviewable.

1

Install

Check the CLI and local environment.

Doctor
2

Initialize

Create or migrate a disposable repository.

Repository
3

Preview

Ask for the change without applying it.

Plan
4

Inspect

Review checks, evidence, and stop reasons.

Result
✓ Reviewable result
Command flow

Make every boundary visible

Connect a plain-language task to a local repository, configured checks, and evidence you can inspect before applying anything.

  • Disposable fixture creation stays local and credential-free.
  • Model-backed preview still requires your provider and entitlement.
  • Checks, retries, and review work stay visible.
bash
# Check environment and initialize the fixture
$ minitok doctor
$ minitok migrate <fixture>
 
# Preview without applying the generated change
$ minitok run --dry-run "Add a health-check endpoint, tests, and documentation."
✓ Plan and checks remain inspectable
Illustrative command flow, not live production evidence.
Run it yourself

Complete procedure

Follow the full path below and keep the output with the disposable repository's change record.

01

Scenario

Task: Add a health-check endpoint, tests, and documentation while preserving existing APIs.

Why this example: it is small enough to reproduce, has a deterministic check, and produces an easy-to-review file and test summary.

02

Install and initialize

Install the CLI and check the environment:

curl -fsSL https://raw.githubusercontent.com/floticinfo/minitok-server/rust-migration/scripts/install-rust.sh | bash
minitok doctor

For an existing disposable repository:

cd your-disposable-repository
git init
minitok migrate
03

Create the local fixture

The promotion helper creates a local Git repository and writes a plan; it does not call a model or apply a change. Pass the printed absolute path to minitok migrate.

# POSIX shell
output="$(minitok promote demo)"
printf '%s\n' "$output"
fixture="$(printf '%s\n' "$output" | sed -n 's/^Demo fixture created: //p')"
minitok migrate "$fixture"

# PowerShell
$output = minitok promote demo
$output
$fixture = (($output | Where-Object { $_ -match '^Demo fixture created: ' }) -replace '^Demo fixture created: ', '').Trim()
minitok migrate $fixture
04

Preview and inspect

A configured provider and entitlement are still required for the model-backed preview:

minitok run --dry-run "Add a health-check endpoint, tests, and documentation. Preserve existing APIs."
minitok status
minitok run list
minitok run show <run-id>

The first local success is a passing verification gate and inspectable evidence record.

What to inspect

  • The plan matches the repository structure and requested compatibility constraint.
  • The repository's VERIFY_CMD.mjs or configured check is the project check.
  • Changed files, test output, retries, and stop reasons are visible.
  • Failed checks stop or guide bounded repair rather than being hidden.

After the preview

If the plan and safety boundaries look right, run the same task without --dry-run only after confirming your provider, entitlement, approval policy, backups, and repository checks.

minitok does not replace Anthropic, OpenAI, Google, or another model provider. Provider credentials and model usage costs are separate.

Evidence boundary: this is a runnable procedure, not a claim that the example has been completed on your repository. A real model-backed run requires a configured provider and an active minitok entitlement. A fixture created by minitok promote demo is onboarding evidence only, not model output or deployment proof. Verified means the configured checks passed; it does not guarantee correctness, security, deployment, or production readiness.

Sources and limits

This walkthrough follows the CLI's documented command shape. See the documentation, AI-assisted repository change case study, source repository, and GitHub Releases. For data handling, see the privacy policy.

Ready to inspect your own change?

Keep the workflow local, bounded, and reviewable.