# Release Upgrade Safety

Audit a pending OpenBot release for upgrade and data-loss hazards before the version is...

- Repository: https://github.com/nightly-labs/openbot (4509 stars)
- Category: Engineering
- License: see repository
- Last updated: 2026-11-01
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add nightly-labs/openbot --skill release-upgrade-safety -g -a claude-code -y`
Codex: `npx skills add nightly-labs/openbot --skill release-upgrade-safety -g -a codex -y`
Cursor: `npx skills add nightly-labs/openbot --skill release-upgrade-safety -g -a cursor -y`
Gemini CLI: `npx skills add nightly-labs/openbot --skill release-upgrade-safety -g -a gemini-cli -y`

## Description

Audit a pending OpenBot release for upgrade and data-loss hazards before the version is bumped or tagged. Use when asked to cut a release, create or bump a version, prepare a tag, or check whether a change is safe to ship to installed users.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Release upgrade safety

Every hazard here is one an installed user pays for and you cannot take back. `openbot.db` is
migrated in place with no backup, two dozen files under `userData` are rewritten by whichever build
opens them last, a released Team API adapter is spoken by peers you will never update, and the
account Worker's D1 migrations are applied before the Worker that needs them. A build that ships
past one of these does not fail on your machine; it fails on someone else's, once, permanently.

Audit the diff since the last released tag against the eight gates below, then report a verdict.
This runs *before* `docs/RELEASING.md`, which stays authoritative for the publish itself.

## What this skill does and does not do

- It audits. It does not run `bun run release:patch`, commit a version bump, or push a tag unless
  you are separately asked to.
- It never runs `bun run check`, `check:desktop`, `test`, or `build-storybook`, each takes minutes,
  CI owns them, and the desktop suite flakes under load, so a red result would tell you nothing.
  Run the narrowest test file named by a gate and lint the changed files. Do not run the repo-wide
  `bun run lint` or `bun run typecheck`: the pre-commit hook and CI run them.
- **If gate C or D fired, confirm the mobile typecheck passed.** `apps/mobile` depends on
  `@openbot/contracts`, which is exactly what those two gates change. The pre-commit hook runs
  `typecheck:mobile` when a commit stages a file that mobile can see, such as one in
  `packages/contracts`, and the CI Surfaces job runs it on each pull request.
- It never runs `bun run dev:seed` or `dev:reset`, both destroy the developer's own profile, and
  never `pkill -f`, which kills other sessions' work mid-write.

## Step 1, establish the range, then account for every file in it

```bash
git status --porcelain
git tag --list 'v*' --sort=-v:refname | head -1
git diff --name-status <tag>..HEAD
```

The rest: https://github.com/nightly-labs/openbot/blob/HEAD/.agents/skills/release-upgrade-safety/SKILL.md