A curated collection of Claude Code skills, agents, and plugins for supercharging your development workflow.
- Overview
- Skills
- Agents
- Plugins
- Installation
- Skill Categories
- Featured Skills
- Usage Examples
- Contributing
- License
ProToolz is a comprehensive collection of professional-grade tools designed to enhance Claude Code's capabilities across multiple domains:
| Domain | Tools | Purpose |
|---|---|---|
| Analysis | technical-explainer | Deep codebase analysis & documentation |
| Design | ui-ux-pro-max | Complete UI/UX design system |
| Visualization | claude-d3js-skill | Interactive D3.js charts & diagrams |
| Testing | cc_chrome_devtools_mcp, playwright-skill | Browser automation & Core Web Vitals |
| Video | remotion-best-practices | React-based video creation |
| Performance | cuda-optimization | GPU/CUDA optimization for LLMs |
| Documentation | compound-docs | Problem-solution documentation |
| Discovery | find-skills | Skill marketplace discovery |
| Memory | chronicler | Institutional memory & logging |
| Workflow | ralph-loop | Iterative AI development loops |
Deep-dive technical explainer for codebases
| Category | Analysis |
| Trigger | Explaining complex systems, codec internals, architectural patterns |
| Output | Interactive web apps, markdown docs, annotated source tours |
Key Features:
- Analyzes actual source code (not just docs)
- Multi-level depth calibration (L1-L5)
- Three visual themes: Light, Academic, Dark
- Side-by-side comparisons
- Data flow visualization
Use for: codec/serialization systems, compiler internals,
ML pipelines, distributed systems
UI/UX design intelligence with 50+ styles & 9 technology stacks
| Category | Design |
| Stacks | React, Next.js, Vue, Svelte, SwiftUI, React Native, Flutter, Tailwind, shadcn/ui |
| Styles | Glassmorphism, Claymorphism, Minimalism, Brutalism, Neumorphism, Bento Grid |
What's Included:
- 50 UI styles with CSS implementations
- 21 color palettes by product type
- 50 font pairings from Google Fonts
- 20 chart recommendations
- 99 UX guidelines with priority ranking
# Generate a complete design system
python3 skills/ui-ux-pro-max/scripts/search.py "saas dashboard modern" --design-systemInteractive data visualizations using D3.js
| Category | Visualization |
| Frameworks | Vanilla JS, React, Vue, Svelte |
| Charts | Bar, Line, Scatter, Pie, Chord, Heatmap, Force-directed networks |
Capabilities:
- Custom chart creation beyond standard libraries
- Network/graph visualizations
- Geographic projections
- Smooth transitions & animations
- Responsive sizing
- Accessibility support
Chrome DevTools automation via MCP
| Category | Testing |
| Protocol | Chrome DevTools Protocol (CDP) |
| Tools | 27 professional-grade testing tools |
Core Web Vitals Support:
- INP (Interaction to Next Paint) ≤ 200ms
- LCP (Largest Contentful Paint) ≤ 2.5s
- CLS (Cumulative Layout Shift) ≤ 0.1
Workflows:
- Performance tracing
- Network monitoring & HAR export
- Accessibility tree inspection
- Device emulation
- Multi-tab management
Complete browser automation with Playwright
| Category | Automation |
| Output | Clean test scripts to /tmp |
| Features | Auto-detects dev servers, responsive testing |
Use Cases:
- E2E testing
- Form validation
- Screenshot capture
- Login flow testing
- Link validation
- Visual regression testing
Video creation in React
| Category | Video |
| Framework | Remotion |
| Topics | 3D, Animations, Audio, Assets, Compositions |
Best practices for building programmatic videos with React components.
CUDA optimization for LLM inference
| Category | Performance |
| Focus | Transformer/LLM runtimes |
| Targets | Atlas, vLLM, ExLlama, Ollama-like systems |
Optimization Areas:
- Attention + KV cache
- GEMM/tensor cores
- Quantization paths (INT4/INT8)
- Kernel fusion
- Launch overhead
- VRAM/fragmentation
Metrics:
- Decode latency (ms/token)
- Prefill throughput (tokens/s)
- VRAM footprint
Problem-solution documentation system
| Category | Documentation |
| Format | YAML frontmatter + Markdown |
| Organization | Category-based directories |
Auto-triggers on:
- "that worked"
- "it's fixed"
- "working now"
Creates searchable institutional knowledge with:
- Exact error messages
- Investigation attempts
- Root cause analysis
- Prevention guidance
- Cross-references
Skill marketplace discovery
| Category | Discovery |
| Triggers | "how do I...", "find a skill for...", "is there a skill that..." |
Helps discover and install additional Claude Code skills from the marketplace.
Institutional memory agent
| Type | Agent |
| Model | Opus |
| Purpose | Record, preserve, and narrate work evolution |
Operating Modes:
| Mode | Use Case |
|---|---|
| Daily Log | End-of-day development activity capture |
| Narrative | Sprint summaries, retrospectives |
| Project Chronicle | Comprehensive project documentation for handoffs |
Sources of Truth:
- Git commits & diffs
- File metadata
- Markdown notes
- Design docs, specs, TODOs
- Issue trackers / PRs
"If there is a conflict between memory and evidence, evidence wins."
Iterative AI development loops
| Type | Plugin |
| Technique | Ralph Wiggum coding methodology |
| Mechanism | Stop hook intercepts exit, feeds same prompt |
How it Works:
/ralph-loop "Build a REST API for todos" --completion-promise "COMPLETE" --max-iterations 50- Claude works on task
- Tries to exit
- Stop hook blocks exit
- Same prompt fed back
- Repeat until completion promise detected
Creates a self-referential feedback loop where each iteration sees modified files and git history.
# Clone the repository
git clone https://github.com/yourusername/ProToolz.git ~/.claude/skills/ProToolz
# Or copy individual skills
cp -r ProToolz/skills/ui-ux-pro-max ~/.claude/skills/~/.claude/skills/
├── technical-explainer/
├── ui-ux-pro-max/
├── claude-d3js-skill/
├── cc_chrome_devtools_mcp/
├── playwright-skill/
├── remotion-best-practices/
├── cuda-optimization/
├── compound-docs/
└── find-skills/
~/.claude/agents/
└── chronicler.md
~/.claude/plugins/
└── ralph-loop/
{
"mcpServers": {
"chrome-devtools": {
"command": "npx",
"args": [
"chrome-devtools-mcp@latest",
"--isolated=true",
"--viewport=1920x1080"
]
}
}
}mindmap
root((ProToolz))
Analysis
technical-explainer
Design
ui-ux-pro-max
Visualization
claude-d3js-skill
Testing
cc_chrome_devtools_mcp
playwright-skill
Video
remotion-best-practices
Performance
cuda-optimization
Documentation
compound-docs
Discovery
find-skills
Memory
chronicler
Workflow
ralph-loop
technical-explainer - Deep Codebase Analysis
Code is the truth. Docs lie.
- Always read source code before writing explanations
- Determine explanation depth based on complexity, not word count
- Show actual code snippets with inline annotations
- Explain the "why" behind design decisions
- Surface trade-offs explicitly
| Level | When to Use | Content |
|---|---|---|
| L1: Glance | Trivial/obvious code | 1-2 sentences, no code |
| L2: Summary | Standard patterns | Purpose + key types + 1 snippet |
| L3: Walkthrough | Important but conventional | Full API + 2-3 snippets + trade-offs |
| L4: Deep Dive | Novel/complex/critical | Everything: algorithm, memory, perf |
| L5: Archaeology | Confusing/surprising | Git blame, commit history, "why?" |
- Interactive Web App (Vite + React)
- Markdown Document
- Annotated Source Tour
ui-ux-pro-max - Design Intelligence
# Generate complete design system
python3 skills/ui-ux-pro-max/scripts/search.py "beauty spa wellness" --design-system
# Search specific domain
python3 skills/ui-ux-pro-max/scripts/search.py "glassmorphism" --domain style
# Get stack-specific guidelines
python3 skills/ui-ux-pro-max/scripts/search.py "responsive forms" --stack html-tailwind| Domain | Use For |
|---|---|
product |
Product type recommendations |
style |
UI styles, colors, effects |
typography |
Font pairings |
color |
Color palettes by product type |
landing |
Page structure, CTA strategies |
chart |
Chart types |
ux |
Best practices, anti-patterns |
- No emojis used as icons (use SVG)
- All clickable elements have
cursor-pointer - Light/dark mode contrast verified
- Responsive at 375px, 768px, 1024px, 1440px
-
prefers-reduced-motionrespected
cuda-optimization - LLM Inference Performance
- Establish baseline - Fixed prompt/seed/config, record metrics
- Classify bottleneck - Bandwidth/Compute/Launch/Occupancy/Layout
- Profile - Nsight Systems (gaps, syncs) → Nsight Compute (memory, occupancy)
- Apply fixes - Highest leverage first
- Verify - Correctness + performance tests
| Type | Symptoms | Key Fixes |
|---|---|---|
| Bandwidth | KV reads, decode attention | Coalesced loads, vectorize |
| Compute | GEMMs, prefill attention | Tensor cores, cuBLASLt |
| Launch | Many tiny kernels, idle gaps | CUDA Graphs, kernel fusion |
| Occupancy | Low SM utilization | Reduce registers/shared memory |
compute-sanitizer --tool memcheck # OOB/illegal
compute-sanitizer --tool racecheck # Races
compute-sanitizer --tool initcheck # UninitializedUser: Explain how this codec works
Assistant: [Uses technical-explainer skill to analyze source code,
generate architecture diagrams, and create annotated documentation]
User: Design a SaaS analytics dashboard
Assistant: [Uses ui-ux-pro-max to generate design system with color palette,
typography, component styles, and accessibility guidelines]
User: Check the Core Web Vitals for my site
Assistant: [Uses cc_chrome_devtools_mcp to run performance traces,
measure INP/LCP/CLS, and provide optimization recommendations]
User: Create an interactive network diagram
Assistant: [Uses claude-d3js-skill to build force-directed graph
with zoom, pan, tooltips, and smooth transitions]
User: My LLM inference is slow at batch=1
Assistant: [Uses cuda-optimization to diagnose decode latency issues,
profile with Nsight, and recommend KV cache optimizations]
User: That worked! The N+1 query is fixed.
Assistant: [Uses compound-docs to capture the solution with YAML frontmatter,
root cause analysis, and prevention guidance]
User: /ralph-loop "Build a complete REST API with tests" --max-iterations 50
Assistant: [Ralph loop iterates, tests, fixes, until completion promise detected]
Contributions are welcome! To add a new skill:
- Create a new directory under
skills/ - Add a
SKILL.mdwith YAML frontmatter:--- name: your-skill-name description: Brief description of what the skill does ---
- Include any reference files in
references/ - Add scripts in
scripts/if applicable - Update this README with skill documentation
MIT License - see LICENSE for details.
Built with Claude Code