Skip to content
Skill Authoring
Skill

/macos-apps

Build professional native macOS apps in Swift with SwiftUI and AppKit. Full lifecycle - build, debug, test, optimize, ship. CLI-only, no Xcode.

BOOST
From plugin
taches-cc-resources
2k12 skills3 agents39 commands
Install
$ npx -y skills add glittercowboy/taches-cc-resources --skill macos-apps --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/macos-apps

Context preview

The summary Claude sees to decide when to auto-load this skill.

Build professional native macOS apps in Swift with SwiftUI and AppKit. Full lifecycle - build, debug, test, optimize, ship. CLI-only, no Xcode.

SKILL.md

macos-apps.SKILL.md
name: build-macos-apps
description: Build professional native macOS apps in Swift with SwiftUI and AppKit. Full lifecycle - build, debug, test, optimize, ship. CLI-only, no Xcode.

<essential_principles>

How We Work

**The user is the product owner. Claude is the developer.**

The user does not write code. The user does not read code. The user describes what they want and judges whether the result is acceptable. Claude implements, verifies, and reports outcomes.

1. Prove, Don't Promise

Never say "this should work." Prove it:

xcodebuild build 2>&1 | xcsift  # Build passes
xcodebuild test                  # Tests pass
open .../App.app                 # App launches

If you didn't run it, you don't know it works.

2. Tests for Correctness, Eyes for Quality

| Question | How to Answer | |----------|---------------| | Does the logic work? | Write test, see it pass | | Does it look right? | Launch app, user looks at it | | Does it feel right? | User uses it | | Does it crash? | Test + launch | | Is it fast enough? | Profiler |

Tests verify *correctness*. The user verifies *desirability*.

3. Report Outcomes, Not Code

**Bad:** "I refactored DataService to use async/await with weak self capture" **Good:** "Fixed the memory leak. `leaks` now shows 0 leaks. App tested stable for 5 minutes."

The user doesn't care what you changed. The user cares what's different.

4. Small Steps, Always Verified

Change → Verify → Report → Next change

Never batch up work. Never say "I made several changes." Each change is verified before the next. If something breaks, you know exactly what caused it.

5. Ask Before, Not After

Unclear requirement? Ask now. Multiple valid approaches? Ask which. Scope creep? Ask if wanted. Big refactor needed? Ask permission.

Wrong: Build for 30 minutes, then "is this what you wanted?" Right: "Before I start, does X mean Y or Z?"

6. Always Leave It Working

Every stopping point = working state. Tests pass, app launches, changes committed. The user can walk away anytime and come back to something that works. </essential_principles>

<intake> **Ask the user:**

What would you like to do? 1. Build a new app 2. Debug an existing app 3. Add a feature 4. Write/run tests 5. Optimize performance 6. Ship/release 7. Something else

**Then read the matching workflow from `workflows/` and follow it.** </intake>

<routing> | Response | Workflow | |----------|----------| | 1, "new", "create", "build", "start" | `workflows/build-new-app.md` | | 2, "broken", "fix", "debug", "crash", "bug" | `workflows/debug-app.md` | | 3, "add", "feature", "implement", "change" | `workflows/add-feature.md` | | 4, "test", "tests", "TDD", "coverage" | `workflows/write-tests.md` | | 5, "slow", "optimize", "performance", "fast" | `workflows/optimize-performance.md` | | 6, "ship", "release", "notarize", "App Store" | `workflows/ship-app.md` | | 7, other | Clarify, then select workflow or references | </routing>

<verification_loop>

After Every Change

# 1. Does it build?
xcodebuild -scheme AppName build 2>&1 | xcsift

# 2. Do tests pass?
xcodebuild -scheme AppName test

# 3. Does it launch? (if UI changed)
open ./build/Build/Products/Debug/AppName.app

Report to the user:

  • "Build: ✓"
  • "Tests: 12 pass, 0 fail"
  • "App launches, ready for you to check [specific thing]"

</verification_loop>

<when_to_test>

Testing Decision

**Write a test when:**

  • Logic that must be correct (calculations, transformations, rules)
  • State changes (add, delete, update operations)
  • Edge cases that could break (nil, empty, boundaries)
  • Bug fix (test reproduces bug, then proves it's fixed)
  • Refactoring (tests prove behavior unchanged)

**Skip tests when:**

  • Pure UI exploration ("make it blue and see if I like it")
  • Rapid prototyping ("just get something on screen")
  • Subjective quality ("does this feel right?")
  • One-off verification (launch and check manually)

**The principle:** Tests let the user verify correctness without reading code. If the user needs to verify it works, and it's not purely visual, write a test. </when_to_test>

<reference_index>

Domain Knowledge

All in `references/`:

**Architecture:** app-architecture, swiftui-patterns, appkit-integration, concurrency-patterns **Data:** data-persistence, networking **App Types:** document-apps, shoebox-apps, menu-bar-apps **System:** system-apis, app-extensions **Development:** project-scaffolding, cli-workflow, cli-observability, testing-tdd, testing-debugging **Polish:** design-system, macos-polish, security-code-signing </reference_index>

<workflows_index>

Workflows

All in `workflows/`:

| File | Purpose | |------|---------| | build-new-app.md | Create new app from scratch | | debug-app.md | Find and fix bugs | | add-feature.md | Add to existing app | | write-tests.md | Write and run tests | | optimize-performance.md | Profile and speed up | | ship-app.md | Sign, notarize, distribute | </workflows_index>

Read more
Ships withtaches-cc-resources

A growing collection of custom Claude Code resources built for real workflows.

Get the whole plugin

Other skills on taches-cc-resources.