jamesrochabrun-skills

Skills Plugin & Marketplace

A comprehensive plugin and marketplace for Claude Code containing 24 custom skills across engineering, Apple development, product management, design, content, trading, database, QA, educational, and AI architecture domains.

Overview

This repository provides Claude Code with specialized knowledge and workflows through a curated collection of skills. Each skill includes interactive scripts, comprehensive reference materials, and best practices for specific domains.

Use as a Marketplace to browse and select individual skills, or install as a Plugin to get all skills at once.

Skills Index

Quick navigation to all 24 skills organized by category:

AI & Architecture (2 skills)

• anthropic-architect - Determine the best Anthropic architecture for your project
• anthropic-prompt-engineer - Master Anthropic's prompt engineering techniques

Engineering (5 skills)

• openai-prompt-engineer - Generate and improve prompts using best practices for OpenAI GPT-5
• git-worktrees - Manage Git worktrees for parallel development
• llm-router - Route requests to different LLM providers
• engineer-expertise-extractor - Extract engineering expertise from GitHub
• engineer-skill-creator - Transform expertise into actionable skills

Apple Development (3 skills)

• swift-concurrency - Build, audit, and refactor Swift 6+ concurrency code
• swiftui-animation - Advanced SwiftUI animations and Metal shaders
• releasing-macos-apps - Create notarized macOS releases with Sparkle updates

Product Management (1 skill)

• prd-generator - Generate Product Requirements Documents

Product Marketing (1 skill)

• technical-launch-planner - Plan technical product launches

Design (4 skills)

• design-brief-generator - Generate comprehensive design briefs
• frontend-designer - Build accessible, responsive UIs
• apple-hig-designer - Design native iOS apps
• book-illustrator - Children's book illustration guide

Content & Writing (2 skills)

• content-brief-generator - Generate SEO-optimized content briefs
• kids-book-writer - Write engaging children's books

Trading & Finance (1 skill)

• trading-plan-generator - Generate trading plans with risk management

Database & Query (1 skill)

• query-expert - Master SQL and database queries

QA & Testing (1 skill)

• qa-test-planner - Generate test plans and bug reports

Educational & Learning (3 skills)

• math-teacher - Interactive math learning with gamification
• reading-teacher - Interactive reading teacher for ages 1-10
• leetcode-teacher - Technical interview preparation

---

Skills

AI & Architecture Skills

anthropic-architect

Determine the best Anthropic architecture for your project. Analyze requirements and receive tailored recommendations on the optimal architecture using Skills, Agents, Subagents, Prompts, and SDK primitives based on proven decision rubrics and 2025 best practices.

Key features:

• Analyze project requirements and complexity
• Recommend optimal architecture (Skills vs Agents vs Prompts vs SDK)
• Apply comprehensive decision rubrics
• Progressive disclosure design guidance
• Security and performance best practices
• Real-world architectural patterns
• Cost optimization strategies

Usage:

• "Using anthropic-architect, determine the best architecture for my code review system"
• "Help me choose between Skills and Agents for this workflow"
• "Analyze my project and recommend the right Anthropic primitives"
• "What's the best architecture for automated testing?"

The Four Anthropic Primitives:

• Skills - Reusable domain expertise with progressive disclosure
• Agents/Subagents - Autonomous task handlers with isolated contexts
• Direct Prompts - Simple, clear instructions for straightforward tasks
• SDK Primitives - Custom workflows and system integrations

Decision Framework:

Task Complexity:
  Low (1-3) → Direct Prompts
  Medium (4-6) → Skills
  High (7-9) → Agents/Subagents
  Custom (10) → SDK Primitives

Plus evaluation of:
- Reusability requirements
- Context management needs
- Security & control requirements
- Performance needs
- Maintenance burden
- Time to market

Reference materials:

• decision_rubric.md - Comprehensive 7-dimension decision framework
• architectural_patterns.md - 14 proven architecture patterns
• best_practices.md - 2025 Anthropic best practices (progressive disclosure, context management, security)
• use_case_examples.md - Real-world architectures for common use cases

Perfect for:

• Choosing the right Anthropic primitives for your project
• Designing scalable, maintainable AI architectures
• Optimizing context usage and costs
• Following 2025 best practices
• Avoiding common anti-patterns
• Learning when to use Skills vs Agents vs Prompts
• Building production-ready AI systems

Example Recommendations:

Use Case: Code Review System

Recommended: Skills + Direct Prompts
Why: Reusable standards (Skills), simple workflow (Prompts)
Pattern: Load code-review-expert skill, review PRs with prompts

Use Case: Full-Stack Feature

Recommended: Specialist Agents + Skills
Why: Complex multi-layer work, parallel execution, specialized expertise
Pattern: Frontend/Backend/DevOps agents each load relevant skills

Use Case: Simple Bug Fix

Recommended: Direct Prompt
Why: One-time task, minimal complexity, no reusability needed
Pattern: Clear instruction to Claude

View anthropic-architect skill →

---

anthropic-prompt-engineer

Master Anthropic's prompt engineering techniques for Claude AI models. Generate new prompts from scratch or improve existing ones using best practices, XML structuring, chain-of-thought reasoning, and progressive disclosure patterns optimized for Claude 4.x.

Key features:

• Generate effective prompts from requirements
• Improve existing prompts with proven techniques
• Claude-specific optimizations (XML tags, thinking, prefilling)
• Apply core techniques: clear instructions, few-shot examples, role assignment
• Advanced patterns: progressive disclosure, chain-of-thought, structured output
• Avoid common mistakes and hallucinations
• Production-ready prompt templates

Usage:

• "Using anthropic-prompt-engineer, create a prompt for extracting structured data"
• "Improve this prompt using Anthropic best practices"
• "Generate a Claude-optimized prompt for code analysis"
• "Debug why this prompt gives inconsistent results"

Core Anthropic Techniques:

• XML Tags - Structure prompts with semantic tags like <instructions>, <example>, <context>
• Chain of Thought - Ask Claude to think step-by-step for complex reasoning
• Prefilling - Start Claude's response to guide format and style
• Few-Shot Examples - Provide 2-5 diverse examples showing desired pattern
• Progressive Disclosure - Load context incrementally to optimize token usage
• Clear Instructions - Claude 4.x excels with explicit, unambiguous direction

Reference materials:

• core_techniques.md - Essential Anthropic techniques with examples
• advanced_techniques.md - Advanced optimization methods
• common_mistakes.md - Pitfalls to avoid
• claude_4_best_practices.md - Claude 4.x specific guidance
• prompt_templates.md - Ready-to-use production templates

Perfect for:

• Creating consistent, high-quality prompts for Claude
• Learning Anthropic-specific techniques (vs OpenAI/GPT patterns)
• Optimizing prompts for Claude 4.x models
• Building prompt libraries for your team
• Reducing hallucinations and errors
• Achieving first-try success more often
• Production prompt engineering

Example Improvements:

Before:

"Analyze this code"

After (Anthropic-optimized):

<instructions>
Analyze the code below for:
1. Security vulnerabilities
2. Performance issues
3. Code quality concerns

For each issue found:
- Specify exact line number
- Explain the problem
- Provide specific fix recommendation
</instructions>

<code>
[code here]
</code>

<output_format>
Return JSON array:
[{
  "type": "security|performance|quality",
  "line": number,
  "issue": "description",
  "fix": "recommendation"
}]
</output_format>

View anthropic-prompt-engineer skill →

---

Engineering Skills

openai-prompt-engineer

Generate and improve prompts using cutting-edge techniques for OpenAI GPT-5 and other LLMs. Apply advanced patterns like chain-of-thought, few-shot prompting, structured output, and progressive disclosure to create effective, consistent prompts.

Key features:

• Generate new prompts from specifications
• Improve existing prompts with proven techniques
• Model-specific optimizations (GPT-5 vs Claude)
• Advanced patterns: Chain-of-Thought, Few-Shot, XML structuring
• Prompt quality analysis and debugging
• Comprehensive best practices library

Usage:

• "Using openai-prompt-engineer, create a prompt for code review"
• "Improve this prompt for better results"
• "Generate a GPT-5 optimized prompt for data analysis"
• "Analyze this prompt and suggest improvements"

Techniques covered:

• GPT-5 specific: Structured prompting (role+task+constraints), reasoning effort control, agentic behavior calibration, verbosity management
• Claude specific: XML tag structuring, step-by-step thinking, progressive disclosure, long-context optimization
• Universal: Chain-of-thought prompting, few-shot learning, zero-shot prompting, self-consistency, role-based prompting

Reference materials:

• gpt5_techniques.md - GPT-5 specific best practices and patterns
• claude_techniques.md - Claude/Anthropic optimization strategies
• prompt_patterns.md - Reusable templates for common tasks
• optimization_strategies.md - Systematic prompt improvement framework

Perfect for:

• Creating consistent, high-quality prompts
• Optimizing prompts for specific models
• Learning advanced prompting techniques
• Reducing API costs through better prompts
• Achieving first-try success more often
• Building prompt libraries for your team

Example improvements:

Before:

"Write code for authentication"

After (GPT-5 optimized):

ROLE: Senior TypeScript developer
TASK: Implement JWT-based authentication service
REQUIREMENTS:
- TypeScript with strict mode
- Include refresh token rotation
- Comprehensive error handling
- Unit tests with 80%+ coverage
OUTPUT: Complete implementation with tests and usage example

View openai-prompt-engineer skill →

---

git-worktrees

Manage Git worktrees for parallel Claude Code development. Run multiple Claude Code sessions simultaneously on different branches without conflicts.

Key features:

• Super simple scripts for worktree management
• Interactive worktree creation and cleanup
• Automatic sync with main branch
• Best practices and command reference
• Custom slash command template

Usage:

• "Create a worktree for feature-api"
• "List all active worktrees"
• "Clean up old worktrees"
• "Sync worktree with main branch"

Includes:

• create_worktree.sh - Interactive worktree creation
• list_worktrees.sh - View all active worktrees
• cleanup_worktrees.sh - Remove old worktrees safely
• sync_worktree.sh - Keep worktrees up-to-date
• Complete command reference and best practices