Reusable AI coding agent skills for Apple platform development.
Works with both Claude Code and Codex.
# Both tools, global (all projects)
./install.sh
# Claude Code only
./install.sh claude
# Codex only
./install.sh codex
# Into a specific project
./install.sh --project ~/workspace/h/InspectThis symlinks each skill into the discovery paths:
- Claude Code:
~/.claude/skills/or<project>/.claude/skills/ - Codex:
~/.agents/skills/or<project>/.agents/skills/
skills/
├── build-debug-ios/ Build, launch, and debug iOS simulator apps
├── build-debug-macos/ Build, launch, and debug macOS apps
├── release-macos/ Archive, notarize, and release macOS apps via GitHub + Homebrew
├── release-testflight-ios/ Archive, export, and upload iOS apps to TestFlight via asc
└── install.sh
Each skill contains:
| File | Used by | Purpose |
|---|---|---|
SKILL.md |
Claude Code | Skill instructions, frontmatter metadata, ${CLAUDE_SKILL_DIR} script refs |
agents/openai.yaml |
Codex | Display name, short description, default prompt |
scripts/ |
Both | Shared helper scripts referenced by either tool |
Full macOS release pipeline: archive → export → notarize → staple → GitHub release → Homebrew cask update. Requires Developer ID signing, notarytool keychain profile, and gh CLI. Manual invocation only (/release-macos <version>).
Build and launch a macOS app, then capture logs and a screenshot for debugging. Uses log stream and screencapture. Outputs artifact paths and a pass/fail verdict.
Full iOS TestFlight pipeline: archive → export IPA → upload to App Store Connect → distribute to TestFlight group. Uses asc CLI for App Store Connect operations. Supports dry-run, custom test notes, and automatic processing wait. Manual invocation only (/release-testflight-ios [version]).
Build and launch an iOS simulator app, then capture logs, a screenshot, and an optional accessibility dump. Uses simctl for simulator control. Supports axe for UI inspection (brew install cameroncooke/axe/axe).