TinyTorch/.claude

📚 TinyTorch .claude Configuration Hub

This directory contains the complete development system for TinyTorch: workflows, agents, and guidelines.

🚀 Quick Start: Slash Commands

Use these commands to trigger workflows:

/website [request]   - Update website content
/module [request]    - Module development  
/test [target]       - Run tests
/tito [request]      - CLI updates
/education [request] - Educational improvements

Example Usage

/website add a debugging guide
/module fix tensor backward pass
/test all checkpoints
/tito add progress visualization
/education improve autograd learning

📁 Directory Structure

.claude/
├── README.md                           # This file (START HERE)
│
├── workflows/                          # 🎯 WORKFLOW DEFINITIONS
│   ├── README.md                      # Workflow system documentation
│   ├── website.md                     # /website command workflow
│   ├── module.md                      # /module command workflow
│   ├── test.md                        # /test command workflow
│   ├── tito.md                        # /tito command workflow
│   └── education.md                   # /education command workflow
│
├── agents/                             # AI AGENT DEFINITIONS (7 Agents)
│   ├── technical-program-manager.md   # Orchestration & coordination
│   ├── education-reviewer.md          # Educational design & validation
│   ├── module-developer.md            # Module implementation
│   ├── quality-assurance.md           # Testing & validation
│   ├── package-manager.md             # Integration & packaging
│   ├── website-manager.md             # Website content & strategy
│   └── devops-tito.md                 # CLI & infrastructure development
│
├── guidelines/                         # DEVELOPMENT STANDARDS
│   ├── DESIGN_PHILOSOPHY.md          # KISS principle
│   ├── GIT_WORKFLOW.md               # Git standards
│   ├── MODULE_DEVELOPMENT.md         # Module patterns
│   ├── TESTING_STANDARDS.md          # Testing requirements
│   ├── PERFORMANCE_CLAIMS.md         # Honest reporting
│   └── AGENT_COORDINATION.md         # Agent teamwork
│
└── docs/                              # Additional documentation

🎯 How the System Works

1. Slash Commands → Workflows

When you type /website add content, Claude:

1. Recognizes the /website command
2. Loads the workflows/website.md specification
3. Executes the defined agent sequence
4. Ensures quality gates are met

2. Workflows → Agents

Each workflow defines:

• Which agents to use
• What order to execute them
• What each agent should produce
• How to handle failures

3. Agents → Implementation

Each agent has:

• Specific expertise area
• Clear responsibilities
• Quality standards
• Handoff protocols

📖 Key Documents to Read

For New Users

1. Start Here: ../CLAUDE.md - Main instructions
2. Workflows: workflows/README.md - How to use slash commands
3. Guidelines: guidelines/DESIGN_PHILOSOPHY.md - Core principles

For Development

• guidelines/GIT_WORKFLOW.md - Git practices
• guidelines/MODULE_DEVELOPMENT.md - Module standards
• guidelines/TESTING_STANDARDS.md - Testing requirements

For Understanding Agents

• agents/ folder - Individual agent capabilities
• workflows/ folder - How agents work together
• guidelines/AGENT_COORDINATION.md - Coordination patterns

🚨 Important Rules

1. Always use slash commands when available
2. Follow workflow sequences - don't skip steps
3. Respect quality gates - fix failures before proceeding
4. Check guidelines before major changes
5. Use version control - work on feature branches

💡 Best Practices

Using Slash Commands

✅ DO: /website add debugging guide ❌ DON'T: "Can you update the website with a debugging guide?"

Following Workflows

✅ DO: Let workflow complete all steps ❌ DON'T: Skip agents or change order

Quality Standards

✅ DO: Fix issues when quality gates fail ❌ DON'T: Bypass testing or validation

🔧 Customization

Adding New Workflows

1. Create workflows/[name].md
2. Define agent sequence
3. Add to slash commands in CLAUDE.md
4. Document in workflows/README.md

Adding New Agents

1. Create agents/[name].md
2. Define capabilities and responsibilities
3. Update relevant workflows
4. Add to agent coordination guide

📊 System Health

Check Configuration

ls -la .claude/          # View structure
ls .claude/workflows/    # List workflows
ls .claude/agents/       # List agents

Validate Setup

• All workflows have corresponding files
• All agents referenced in workflows exist
• Guidelines are up to date
• No orphaned or duplicate files

---

Questions? Start with workflows/README.md for detailed workflow documentation.