9.7 KiB
Release checklist
Run this before every release of the desktop app, on each system the release builds for: Windows x64, macOS Apple silicon and Linux x64. Automated tests cover the logic and never start a real window, a real terminal panel, or a real Claude, Devin or Codex, so everything here is by hand.
Work from the pre-release build: publish a GitHub pre-release so release-app.yml attaches the files,
then install from it. Tick each box per system and note anything that is a rough edge rather than a
failure.
npm run build here runs tauri build with bundling on, so a local build also produces .msi/NSIS
.exe/.dmg/.deb installer packages under src-tauri/target/release/bundle/. Ignore those: the
release workflow builds with --no-bundle and attaches only the plain executable per system (see
.github/workflows/release-app.yml); install.js is the only installer, and it places that plain
file directly after checking it against the release's checksum list.
Before you start
node scripts/check-versions.mjs --tag v<version>passes (version.json, package.json, CHANGELOG.md and the app's tauri.conf.json, package.json and Cargo.toml all agree)npm testpasses at the repo root, with the app absentcargo fmt --check,cargo clippy --workspace --all-targets -- -D warningsandcargo test --workspacepass inplan2code-app/, and theappworkflow is green on all three systems- The pre-release has one
plan2code-app-<version>-<os>-<arch>file per system and aplan2code-app-<version>-SHA256SUMS.txt
Install and launch (each system)
- Run the installer from the pre-release: it offers the app, downloads it, verifies the checksum
and places it in
~/.plan2code/app/ - A deliberately corrupted download is rejected with the checksum message, and the skills and
the
plan2codecommand are still installed plan2codein a project asks for the agent and model, then opens the app on that folder- The agent and model picked in the launcher are the ones the app starts at once, with no launch card or checklist shown first, and New Session keeps them
plan2code --classicruns the agent in the terminal with the app installed- Linux only: with
DISPLAYandWAYLAND_DISPLAYunset,plan2coderuns the terminal flow - The desktop shortcut opens the folder picker and then the app
- macOS: the first start is stopped by Gatekeeper, Open Anyway in System Settings (Privacy &
Security) lets it through, and
docs/desktop-app.mddescribes what you saw - macOS and Linux: start the app from the Dock, Finder or a
.desktopfile (not a terminal) on a machine where Node is only on PATH through nvm, Homebrew or Volta: the first-run checklist still finds Node and the installed CLIs, and the console and an agent still start - Windows: SmartScreen's More info, then Run anyway, lets it through, and
docs/desktop-app.mddescribes what you saw plan2code updateupdates the skills, the command and the app, including while the app is running (Windows replaces the file by renaming the old one aside)
The window (each system)
- The dashboard shows in the frame and the terminal panel is visible beneath it
- Resizing the window keeps both usable; dragging the handle resizes the panel
- The bar shows the project folder; the permission note appears only while an agent started with checks off is running, never otherwise
- The OS window title (and so the taskbar entry) reads Plan2Code - <folder name>, not just Plan2Code; open a second window on a different folder and confirm they read apart in the taskbar without needing to restore either one
- The first run (no
firstRunDonesaved yet) shows the checklist (Node, an agent CLI, the project folder, the skills version) before the launch card; a later launch goes straight to the launch card - The launch card offers agent, a model dropdown (the same models.json list the
plan2codelauncher asks with, hidden for an agent with no listed models), effort (Claude and Codex only) and permission mode (checks off is ticked by default, matching the plainplan2codelauncher; untick it for an agent that keeps its normal approval prompts); a CLI it could not find is named, not just disabled - Dock the terminal to the right: the splitter becomes vertical, the toolbar stacks, and the dashboard keeps the rest of the width; dock it back to the bottom
- Collapse the terminal to a thin strip (either dock) and View Agent Terminal brings it back; its size from before collapsing is remembered
- Docked right, collapsed: the View Agent Terminal label reads top-to-bottom and fits inside the strip, not sideways text running off the edge
- Collapse the terminal (or shrink it small), let the agent go quiet for a while mid-answer: the "agent may need you" badge appears on the strip; expanding it clears the badge
- The top bar's project and agent menus open under their name, close on a click elsewhere (the console frame and the terminal included) or Escape, and are not cut off by the bar
- Change project in the top bar's project menu, with an agent running: it asks first, then the folder picker opens; canceling the picker leaves the running agent and its project untouched
- Change project, pick a different folder: the old agent and console stop, the new folder opens on the launch card (or the checklist, if its own first run has not passed yet)
- Reload the window (or navigate away and back) while an agent is running: the terminal redraws from where it was, not blank, and the dashboard reconnects to the same session
- When the app and the installed skills differ in version, the bar's right end offers Update skills, and nothing updates until it is pressed; with the console stopped as well, Reconnect the console shows first and "+1 more" names the update
The terminal panel, with each real agent (each system)
Do each of Claude Code, Devin and Codex for a few minutes:
- Startup screen draws correctly
- Typing a prompt and getting an answer
- A long streamed answer scrolls and stays readable
- Resizing the window reflows the agent's screen
- Pasting several lines
- A line with Unicode and emoji
- Select text in the terminal: it copies on its own; Ctrl+Shift+C (Windows, Linux) or Cmd+C (macOS) copies the selection too; Ctrl+V / Cmd+V pastes
- Exiting the agent: the agent menu reads "ended" with the reason beside it, and New Session stays just left of it
Agent lifecycle
- Start, then New Session: it confirms once, with Planny on a coffee break and no claim about unsaved work, then a fresh agent starts on the same agent, model, effort and permission mode, and the dashboard shows the pick-up chip with a brief that matches the session
- An agent that ends by itself leaves the app asking nothing until New Session or Change agent (in the agent menu) is pressed; Change agent reopens the launch card with a real choice to make again
- Closing the window while an agent runs asks first, and "no" keeps the window open
- Cmd+Q on macOS asks the same
- Windows only: start a CLI installed through npm (a
.cmdshim), then New Session; check Task Manager and confirm no leftovernode.exe(or the CLI's own process) is left running
Agents and the taskbar (each system)
- A custom agent added in the console's User Preferences → Agents & models (the fake agent in
fake-agent/works) shows in the launch card's picker after the built-ins, starts, and New Session starts the same agent again; the bypass box adds its bypass arguments - With no agent CLI on
PATHand no custom agent, the first-run checklist and the launch card show the explainer with each install command and an Install page link, and Start is off - Each Install page link opens the right page in the default browser (Claude Code, Devin, Codex)
- Install an agent (or add a custom one), press Check again: it appears and Start turns on without restarting the app
- With only one agent installed and nothing remembered, the launch card starts on that agent, not Claude Code
- Windows: the taskbar button shows a green dot when it is your move and a red dot when Plan2Code needs attention, a green dot on a finished screen that still offers a button (back to the dashboard, a review), and no dot while the agent is busy, once a finished step has nothing left to press, or after the agent ends
- Windows: with the window unfocused, the button gives one short burst of flashes when it becomes your move or needs attention (not again on a repeat) and stays lit until the window gets focus
- macOS: the dock shows a red
!badge when Plan2Code needs attention (no badge on your move), and bounces once when it becomes your move or needs attention while the app is in the background - Linux: where the desktop shows badges, a count of 1 appears on needs attention, and the window is marked as wanting attention while unfocused
- Reloading the window, or ending the agent, clears the dot
- User Preferences → System → Desktop app (shown only inside the app, not in a browser tab) has Show on the taskbar when Plan2Code needs me, on by default; switching it off clears the dot at once and stops flashes, and the choice is still off after quitting and reopening the app
Standalone path
/plan2codetyped inside an agent with the app closed opens the dashboard exactly as beforenpm teststill passes with the app installed on this machine