agentsclimarketplace

Desktop app

Skill Skryx-L-A/project-kit/skills/desktop-app

Say "new project" → get a perfectly-prepared project folder. A Claude Code bootstrap kit that grills the plan to a Definition of Ready, then auto-scaffolds files, memory, project sub-agents & tooling — routing to type-specific sub-skills (website, api, data/ml, quant, SaaS, CLI, app, game-mod, research, OSS… + a 7-day build-business ultraskill).

Install
npx -y skills add Skryx-L-A/project-kit --skill desktop-app

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Build and set up a desktop / cross-platform application project (Electron, Qt/Python, or Tauri) with per-OS packaging, code signing, auto-update, and a GitHub-Release-plus-download-page distribution flow. This is a project-kit sub-skill loaded by new-project routing. Use WHENEVER the user wants to build, create, or ship a desktop app, a native GUI, a tray/menu-bar utility, a cross-platform installer, or anything that runs as an installed program on Windows / macOS / Linux.

SKILL.md

6.8 KB, as published. Nobody here has run it

desktop-app — installed cross-platform applications

This sub-skill stands up a desktop application that users download and install on Windows, macOS, and/or Linux. It mirrors the proven release flow of a cross-platform Python+Qt desktop app with local ML weights (e.g. a local-speech tool): Python + Qt control center, local model weights, packaged online + offline installers, GitHub Releases, a Netlify download site, big assets on Hugging Face, Windows build on a remote host, portable offline build in a container. Distribution is treated as a first-class deliverable, not an afterthought.


What this sub-skill is for

A program that runs as an installed binary with a real OS presence (window, tray/menu-bar, notifications, file associations), shipped as signed per-OS installers. Use it for GUI tools, utilities, and tray apps — not for web UIs (website), pure command-line tools (cli-tool), or mobile (mobile-app), though it composes with cli-tool when the app also ships a launcher.

Mandatory grill-questions (fold into the DoR)

  1. Framework — Electron (web stack, biggest bundle), Qt/Python (native feel, PySide6/PyQt), or Tauri (Rust core, tiny bundle, system webview). Recommended: Qt/Python for a native feel unless web UI reuse forces Electron. Lock this before any code.
  2. Target OSes — which of Windows / macOS / Linux, and the minimum versions. Each adds a packaging + signing + CI lane.
  3. Code signing — do you have (or will you get) an Apple Developer ID + notarization, and a Windows Authenticode cert? If not, ship unsigned and warn users of the OS gatekeeper prompts.
  4. Auto-update — needed? (electron-updater / Tauri updater / Sparkle / custom check against GitHub Releases). Define the update channel and the signing key for update artifacts.
  5. Offline / heavy assets — large models or data (like bundled ML weights)? Decide online-installer-fetches-on-first-run vs. offline multi-part bundle, and where big files live (Hugging Face / GitHub Release assets / CDN).
  6. Distribution surface — GitHub Release + a download page (Netlify)? Confirm the download-site repo and the artifact naming scheme.
  7. Build hosts — can all target OSes be built here, or is a remote build host needed (e.g. SSH to a Windows machine for the Windows lane)? Flag any cross-build gaps now.
  8. Data & privacy — local-only vs. cloud; where user data/config lives per-OS; any telemetry (default: none, opt-in only).

Project sub-agents to generate (<project>/.claude/agents/)

  • packager (delegate-by-default) — produces per-OS installers (NSIS/MSI, dmg/pkg, AppImage/deb/rpm or Tauri bundles); owns the build matrix and artifact naming.
  • release-engineer (delegate-by-default) — cuts the GitHub Release, uploads artifacts + checksums, syncs the download page, pushes big assets to Hugging Face; never invents version tags.
  • installer-tester — installs each artifact on a clean target, verifies launch, signing, auto-update, and offline-asset fetch; reports honestly when a lane is broken.
  • ui-reviewer — checks the GUI against the done-bar and the no-emoji rule (typographic symbols / drawn elements only in UI and notifications).
  • Plus the kit defaults: reviewer (diff/PR) and verifier (launches the app to confirm behaviour).

Tools / CLIs / MCP / skills needed

  • Runtime/build: Node + the framework toolchain (electron-builder / @tauri-apps/cli / Rust+cargo) or Python + PySide6/PyQt + PyInstaller/briefcase. Check in environment-readiness.
  • Signing/notary: codesign/notarytool (macOS), signtool/osslsigncode (Windows) — only if signing was chosen; flag missing certs, never fabricate credentials.
  • Distribution: gh (GitHub Releases), netlify-cli (download page), huggingface_hub / git-lfs for big assets. Container runtime (Docker/Podman) for the portable offline build (glibc-pinned) and remote ssh to a build host for OS lanes that can't build locally.
  • CHAIN global skills: cli-anything (if wrapping an existing GUI into an agent-drivable CLI), verify (run the built app), code-review, doc-coauthoring (README + install docs), update-config (permissions/hooks for build commands). For the desktop UI itself, the framer-inspiration / design-harvest skills inform layout even though it isn't a website.

File / asset nudges (on top of the base set)

  • packaging/ — per-OS build configs (electron-builder.yml / tauri.conf.json / PyInstaller specs).
  • packaging/signing/ — signing notes + a .gitignored slot for certs/secrets (never committed).
  • .github/workflows/release.yml — the build matrix + release pipeline.
  • download-site/ (or a linked repo) — the download page.
  • assets/ + an ASSETS.md documenting where heavy files live and how the online installer fetches them.
  • INSTALL.md — per-OS install, the unsigned-app gatekeeper steps, and uninstall.

Stack defaults & done-bar

Default stack: Qt/Python (PySide6) + PyInstaller, signed where certs exist, auto-update checked against GitHub Releases, online installer + optional offline bundle, download page on Netlify, big assets on Hugging Face — i.e. the Python+Qt local-weights pattern. (Switch to Electron/Tauri if grilled.) Done-bar (all true): a signed (or knowingly-unsigned) installer for each target OS installs on a clean machine; the app launches, shows its window/tray, and survives a restart; auto-update applies a newer release end-to-end; offline assets resolve on first run; a GitHub Release exists with artifacts + checksums and the download page links to them; INSTALL.md is accurate.

Guardrails

  • No emojis in the app UI or notifications — typographic symbols or drawn elements only (standing user rule).
  • Honesty about signing: if the app is unsigned, say so plainly and document the OS warnings — never imply it's trusted.
  • Never commit certs, signing keys, notary creds, or API tokens.gitignore them; ask the user to supply them.
  • Don't claim a build lane works until installer-tester proves it on that OS; report cross-build gaps instead of faking them.
  • Commit under the user's own name only (Skryx-L-A) — never add Claude as a co-author.
  • Verify big-asset licensing before redistributing models/data; record provenance in ASSETS.md.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.