YouTube to Tutorial: 10 Claude Code Tricks

A case study in converting a 12-minute YouTube video into a comprehensive, self-contained developer tutorial — using Claude Code itself as the conversion tool.

What This Is

This repository demonstrates a repeatable pipeline for turning video content into actionable written documentation. The source material is an AI Labs video covering 10 Claude Code power tricks. The output is a 1,750-line tutorial that a developer with zero Claude Code experience can follow from start to finish.

The 10 Tricks

#	Trick	What It Does
1	/insights	Analyzes past sessions, generates improvement reports
2	CLAUDE.md Context	4 foundational docs (PRD, architecture, decisions, feature tracking) for rich project context
3	Context7 MCP	Live library/framework documentation via MCP
4	Hooks + Exit Code 2	Shell hooks that control agent behavior with blocking errors
5	Test Protection	Blocks test file modifications during TDD
6	MCP Tool Search	On-demand MCP tool loading to save context window space
7	Git Worktrees	Parallel isolated sessions on separate features
8	Strict Mode	Catch bugs at compile time (TypeScript, Python, Go, Rust)
9	Adversarial Agents	Multi-agent fact-checking and code review
10	Browser Verification	Accessibility tree-based UI testing with 95%+ token savings

Why This Exists

Video tutorials have a fundamental limitation: information is locked in a format that can't be searched, copied, or version-controlled. A 60-second video segment becomes 150-200 lines of written tutorial because written formats require:

• Complete setup steps (videos assume you'll figure it out)
• Full config file contents (videos show a brief flash on screen)
• Multiple examples (videos typically show one)
• Troubleshooting (videos don't cover what happens when things break)
• Platform-specific notes (videos show one OS)

This project solves that by extracting the knowledge from the video and repackaging it as a self-contained reference.

Repository Structure

.
├── README.md                              # This file
├── .claude/
│   ├── settings.json                      # Claude Code config (agent teams enabled)
│   └── skills/
│       └── transcript-to-tutorial/        # Reusable skill for video-to-tutorial conversion
│           ├── SKILL.md                   # 7-phase pipeline skill definition
│           └── tutorial-template.md       # Output template for generated tutorials
├── docs/
│   ├── 10_claude_code_tricks_tutorial.md   # The tutorial (main deliverable)
│   ├── conversion_process.md              # How the conversion was done
│   ├── README_transcript.txt              # Raw video transcript + metadata
│   └── gemini3_summary.txt               # AI-generated structured summary
└── .gitignore

The Conversion Pipeline

The tutorial was produced through a 7-phase pipeline, fully documented in docs/conversion_process.md:

  Video Transcript          AI Summary
       │                        │
       └──────────┬─────────────┘
                  │
          ┌───────▼────────┐
          │  Cross-reference │
          │  against official │
          │  Claude Code docs │
          └───────┬────────┘
                  │
          ┌───────▼────────┐
          │  Plan tutorial   │
          │  structure with  │
          │  parallel agents │
          └───────┬────────┘
                  │
          ┌───────▼────────┐
          │  Write complete  │
          │  tutorial in one │
          │  pass            │
          └───────┬────────┘
                  │
          ┌───────▼────────┐
          │  Automated       │
          │  review (9/10)   │
          └───────┬────────┘
                  │
          ┌───────▼────────┐
          │  Fix issues,     │
          │  commit          │
          └────────────────┘

Key Insight: Fact-Checking Caught Outdated Information

The video referenced an MCP_CLI experimental flag for reducing context bloat. Cross-referencing against the official docs revealed this had been replaced by ENABLE_TOOL_SEARCH, enabled by default since v2.1.7. Without the fact-checking phase, the tutorial would have taught a deprecated approach.

Quick Start

Read the tutorial: Open docs/10_claude_code_tricks_tutorial.md and start from the Prerequisites section.

Understand the process: Read docs/conversion_process.md for the full methodology.

Replicate for your own video: The conversion process doc includes a step-by-step template in its final section.

Use the Claude Code skill: Run /transcript-to-tutorial <path-to-transcript> in Claude Code to automate the pipeline. The skill is at .claude/skills/transcript-to-tutorial/SKILL.md.

What Makes the Tutorial Self-Contained

Every trick includes:

• Setup: Dependencies, configs, and commands — nothing left to guess
• Invocation: Exact steps to trigger the trick inside Claude Code CLI
• Examples: 2-3 practical scenarios with complete, copy-pasteable code
• Verification: How to confirm the trick is working
• Tips: Power-user advice and platform-specific gotchas

Supporting sections:

• Prerequisites: All dependencies consolidated in one place
• Cheat Sheet: One-row-per-trick quick reference table
• Troubleshooting: 7 common issues with fixes
• Glossary: 15 key terms defined

Human Effort

~30 minutes of oversight across two Claude Code sessions:

1. Provide the initial prompt with source material
2. Approve the plan
3. Request commit
4. Request process documentation
5. Request this README

Everything else — research, fact-checking, planning, writing, reviewing, fixing — was autonomous.

License

This project is provided as-is for educational purposes. The source video is by AI Labs. Claude Code is by Anthropic.