Boilerplate upgrades

Ce contenu n’est pas encore disponible dans votre langue.

A project created from this boilerplate diverges from day one. Your routes, models, and business logic are yours — so when the boilerplate later improves a lint rule, an auth pattern, or a CI step, you can’t merge that improvement as a plain git diff. The diff would land on top of code that has moved on.

Boilerstone (the .boilerstone/ directory) is how we solve that. Each boilerplate release publishes migration intentions: short markdown files that say what changed, why, and how to tell whether it concerns your project. You — or an AI agent you supervise — replay the smallest equivalent change in your own code. Nothing is applied automatically.

What you need to know as a consumer

Onboard once — if the project isn’t tracked yet:

curl -fsSL https://raw.githubusercontent.com/lonestone/lonestone-boilerplate/main/install.sh \
  | sh -s -- onboard

Stage an upgrade — this creates a dedicated branch and a disposable workspace. It does not edit your app code:
Terminal window
```
pnpm boilerplate upgrade status
pnpm boilerplate upgrade
```
Apply intentions — one at a time, with a human confirming apply/skip. Follow the upgrade runbook. The Claude Code / Cursor boilerstone-upgrade skills follow that same runbook.
Finish — when every staged intention is applied or skipped:
Terminal window
```
pnpm boilerplate upgrade finish --to <version>
```

For the why behind the system and what each command does, read how it works.

Design choices in one breath

Ship meaning, not diffs — the consumer’s code has moved on.
Everything is markdown and JSON — same artifacts for a human or any agent.
The CLI stages work; it never edits application code.
Leaving is cheap: rm -rf .boilerstone and drop the boilerplate script.