swift-binding-assistant

Claude Skills

Skills for Claude Code.

Installation

Install via the Claude Code plugin system:

/plugin marketplace add justinwojo/claude-skills
/plugin install swift-binding-assistant@justinwojo-claude-skills

Or manually copy individual skills into .claude/skills/ (project-scoped) or ~/.claude/skills/ (global).

Skills

swift-binding-assistant

Create .NET C# bindings for Swift and Objective-C Apple platform libraries — go from an SPM package URL or xcframework to a ready-to-use NuGet package.

Built on top of swift-dotnet-bindings and spm-to-xcframework.

Supported platforms: iOS, macOS, Mac Catalyst, tvOS

Supported framework types: Swift, Objective-C, and mixed — auto-detected during build.

Workflow:

SPM package URL / xcframework / GitHub release
  → Check prerequisites (macOS, Xcode, .NET 10)
  → Build xcframework from SPM (if needed)
  → Create binding project (dotnet new swift-binding)
  → Build & diagnose errors (auto-fetches latest troubleshooting docs)
  → Package as NuGet (.nupkg)
  → Optional: review generated binding for completeness

Example usage:

"Bind the Nuke Swift library for iOS"

"Create a C# binding for this xcframework at ~/frameworks/MyLib.xcframework"

"I have a binding project that's failing to build, help me fix it"

Prerequisites: macOS with Xcode and .NET 10 SDK. No additional configuration needed.

See swift-binding-assistant/SKILL.md for the full skill specification.

---

ai-pair-programming

Query external LLMs (OpenAI/ChatGPT, Google Gemini, xAI Grok) for code review, improvement suggestions, and collaborative problem-solving.

Supported providers:

• OpenAI (gpt-5.4, gpt-5, gpt-4o, o3-mini)
• Google Gemini (gemini-3.1-pro-preview, gemini-3-flash-preview, gemini-2.5-flash)
• xAI Grok (grok-4-1-fast-reasoning, grok-4-1-fast-non-reasoning)

Setup: Set API key environment variables for the providers you want to use: OPENAI_API_KEY, GOOGLE_AI_API_KEY, XAI_API_KEY.

Optionally override default models: OPENAI_MODEL, GEMINI_MODEL, GROK_MODEL.

See ai-pair-programming/SKILL.md for full documentation.

---

session-orchestrator

Autonomously execute multi-session design docs using agent teams. A lead orchestrator reads your design doc, spawns one worker per session sequentially (each with fresh context), verifies deliverables, and reports results.

How it works:

Lead (reads design doc, creates team, directs everything)
 ├── session-1-worker → implements, validates, reviews, commits
 ├── session-2-worker → picks up from session 1, same workflow
 └── final validation, cleanup, summary to user

Setup: Requires the experimental agent teams feature. Add to ~/.claude/settings.json:

{
  "env": {
    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
  }
}

If the ai-pair-programming skill is installed and configured, workers automatically get external AI code review before committing. Without it, workers perform a self-review instead.

Stuck detection: If you use iTerm2 with the it2 CLI and Python API enabled, the orchestrator automatically monitors worker terminal sessions via /loop and intervenes when a worker hangs — sending Ctrl+C and escalating to "resume" nudges as needed. Without iTerm2, the lead still receives teammate messages and idle notifications for basic visibility.

Example usage:

"Orchestrate this design doc: docs/refactor-plan.md"

"Run the sessions from my design doc"

See session-orchestrator/SKILL.md for the full skill specification.

---

smart-permissions

A PreToolUse hook that replaces the built-in permission prompts with configurable, rule-based auto-approval — safe commands execute immediately, dangerous patterns are blocked, and everything else falls through to an optional LLM safety evaluation or the standard permission prompt (with "Always allow" support).

Includes MCP tool support with glob patterns for granular read-only vs write approval (e.g., mcp__sentry__get_*).

Setup: Install the plugin — no API keys required. Optionally set SAFETY_HOOK_API_KEY (plus SAFETY_HOOK_API_URL and SAFETY_HOOK_MODEL) for LLM fallback evaluation. Customize allowed commands/paths/MCP tools via ~/.claude/smart-permissions-config.json.

See smart-permissions/README.md for full documentation.

License

MIT