chore(repo): version packages

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@shelve/cli@5.1.0

Minor Changes

• #751 f9a0848 Thanks @HugoRCD! - Add agent-friendly global flags (--json, --quiet, --yes, --non-interactive), structured JSON errors on stderr, non-interactive guards across all commands, and machine-readable output for query commands. Adds shelve doctor, run spawn JSON events, GitHub Action, CI playground smoke tests, LP skills manifest check, troubleshooting docs, shelve-app agent skill, and llms.txt skill links. Includes login --token, generate --type, HTTP debug logging, fixes silent fetch failures in shelve run, full CLI documentation on shelve.cloud, and a Docus-published agent skill at /.well-known/skills/.

• #752 787cf07 Thanks @HugoRCD! - Add sync policies for push/pull conflict handling, shelve diff and shelve sync, server-side protected environments on projects, and consolidate published agent skills into a single comprehensive shelve skill (remove shelve-app).

Patch Changes

• #749 e64d2ec Thanks @HugoRCD! - Add a self-contained playground/run/ fixture and root pnpm play scripts so the full shelve run flow can be exercised end-to-end without leaving the repo, hitting the production API, or running the Shelve app.

  • A tiny zero-dep node:http fake Shelve API (server.mjs) seeded with a team / project / 3 environments / variables / a fake token (seed.json). Implements the four endpoints shelve run calls.
  • An orchestrator (start.mjs) that boots the server, waits for /health, then spawns the locally-stubbed CLI with the right SHELVE_* env vars injected and tears the server down on exit / Ctrl-C.
  • Root scripts: pnpm play (= shelve run dev), pnpm play:start, pnpm play:fail, pnpm play:watch, pnpm play:server (server only). Each runs pnpm -C packages/cli stub first so edits in packages/cli/src/** are picked up without a rebuild.
  • Two admin endpoints (POST /__playground/variables, POST /__playground/reset) let you trigger watch-mode reloads from another terminal with a single curl.
  • Not part of the pnpm workspace — pnpm install ignores it, no host-project lockfile drift can swallow output.

  Also fixes several long-standing shelve run correctness and cross-platform bugs the playground surfaced:

  • CLI orphaned its own child. runMain(main).then(() => process.exit(0)) resolved as soon as the run command function returned (right after spawn), killing the CLI before the child had done anything. With stdio: 'inherit' the orphaned child kept writing to the terminal so it looked fine, but signal forwarding, watch mode and exit-code propagation were all broken. Fix: the run command awaits a never-resolving promise; child.on('exit') is now the only path to process.exit.
  • --watch --restart-on-change killed the CLI on the first reload. Killing the old child fired its exit handler which called process.exit(143) before the new child could spawn. Fix: mark the outgoing child with __restarting so its exit handler is a no-op while watch is swapping in a replacement.
  • Signal handler leak in watch mode. Each respawn registered fresh process.on(SIGINT|TERM|HUP) listeners. After 10 reloads Node warned. Fix: register process-level signal handlers exactly once and have them target the current child via a module-level ref.
  • process.kill(-pid, …) doesn't work on Windows. Replaced everywhere with tree-kill, which uses taskkill /F /T on Windows and process-group kill on Unix. Dropped detached: true on spawn — no longer needed for cleanup and was what made the CLI miss SIGHUP when the terminal closed.
  • Package-manager shims (pnpm.cmd, npm.cmd, yarn.cmd) didn't spawn on Windows. Node's spawn doesn't resolve .cmd without a shell. Now uses shell: true on Windows (the shim chain is killed correctly thanks to tree-kill).
  • Parent-death watchdog. Both the CLI's run command and the playground orchestrator poll their original ppid with process.kill(ppid, 0) every 2s. If the parent disappears (terminal force-quit, parent SIGKILL'd, Cursor crashed, …), the child tears down its own subtree instead of running orphaned forever and spamming setRawMode EIO into whatever terminal it ended up attached to.
  • Suspiciously-fast exit warning. If a spawned child exits cleanly in <250ms (probably script-resolution went wrong), the CLI now prints a warning pointing at --debug and shelve run -- <pm> <script> to bypass script resolution. Script-resolution errors are debug-logged instead of swallowed.

  Smoke-tested all variants — SIGINT, terminal SIGHUP, parent kill -9, watch+restart, failing exit codes — on macOS. No orphans, ports always released, exit codes always propagated.

@shelve/app@3.2.0

Minor Changes

• #752 787cf07 Thanks @HugoRCD! - Add sync policies for push/pull conflict handling, shelve diff and shelve sync, server-side protected environments on projects, and consolidate published agent skills into a single comprehensive shelve skill (remove shelve-app).

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
shelve-app	Ready	Preview, Comment, Open in v0	May 30, 2026 7:44pm
shelve-lp	Ready	Preview, Comment	May 30, 2026 7:44pm
shelve-vault	Ready	Preview, Comment	May 30, 2026 7:44pm

[vercel]

Deployment failed with the following error:

There is no GitHub account connected to this Vercel account.