ChernyCode

A template repository implementing Boris Cherny's productivity tips for AI-assisted coding with Claude Code and Cursor.

Boris Cherny is the creator of Claude Code. This repo synthesizes his recommendations from two threads on how he and the Claude Code team use the tool, providing ready-to-use configurations for maximum productivity.

What's Included

File/Directory	Purpose	Install To
CLAUDE.md	Project memory for Claude Code	Project root
AGENTS.md	Agent instructions for Cursor	Project root
.cursor/skills/	Project-specific Cursor skills	Project .cursor/skills/
claude_personal_skills/	Personal Claude Code skills	~/.claude/skills/
claude_subagents/	Claude Code subagent definitions	~/.claude/agents/
cursor_personal_skills/	Personal Cursor skills	~/.cursor/skills-cursor/
cursor_subagents/	Cursor subagent definitions	~/.cursor/agents/

Quick Start

1. Copy Project Files

Copy these files to your project root:

• CLAUDE.md - Edit for your project's context
• AGENTS.md - Edit for your project's context
• .cursor/skills/ - Copy the skills directory

2. Install Personal Files

Copy these from this repo to your home directory for use across all projects:

# Create directories if they don't exist
mkdir -p ~/.claude/skills ~/.claude/agents ~/.cursor/skills-cursor ~/.cursor/agents

# Claude Code personal skills
cp -r claude_personal_skills/* ~/.claude/skills/

# Claude Code subagents
cp -r claude_subagents/* ~/.claude/agents/

# Cursor personal skills
cp -r cursor_personal_skills/* ~/.cursor/skills-cursor/

# Cursor subagents
cp -r cursor_subagents/* ~/.cursor/agents/

3. (Optional) Install Personal CLAUDE.md

Create a personal memory file for all projects:

# Copy the template (edit with your preferences)
cp CLAUDE.md ~/.claude/CLAUDE.md

Key Concepts

Memory Files (CLAUDE.md / AGENTS.md)

These files provide persistent context that the AI reads at the start of every session:

• Project-level (./CLAUDE.md, ./AGENTS.md): Shared with your team via git
• Personal (~/.claude/CLAUDE.md): Your preferences across all projects
• Local (./CLAUDE.local.md): Personal project settings, gitignored

Best Practice: After every correction, say:

"Update CLAUDE.md so you don't make that mistake again"

Claude is excellent at writing rules for itself.

Skills

Skills are reusable workflows you can invoke with /skill-name:

Skill	Description
/commit-push-pr	Commit, push, and create a PR
/techdebt	Find and fix technical debt
/code-simplifier	Clean up code after changes
/code-style	Python code style guidelines
/testing	pytest conventions
/git-workflow	Git and commit conventions
/llm-development	LLM/ML best practices

Subagents

Subagents run specialized tasks in their own context window, enabling parallel execution and context isolation:

Agent	Description
code-reviewer	Review code as a senior engineer (readonly)
test-writer	Write comprehensive tests (proactive)
doc-generator	Generate documentation
verifier	Validate completed work (fast model)

Use subagents by:

• Slash command: /code-reviewer review my changes
• Natural language: "use the verifier agent to confirm the feature is complete"

Boris Cherny's Top Tips

1. Start in Plan Mode

For complex tasks, start in Plan mode:

• Cursor: Toggle plan mode in the UI
• Claude Code: Press Shift+Tab twice

Pour your energy into the plan. A good plan lets Claude one-shot the implementation.

2. Work in Parallel

Run multiple Claude sessions simultaneously:

• Use git worktrees for parallel branches
• Run 3-5 sessions on different tasks
• Hand off sessions between terminal and web

3. Create Skills for Repeated Workflows

If you do something more than once a day, make it a skill:

# Create a new skill
mkdir -p ~/.claude/skills/my-skill
cat > ~/.claude/skills/my-skill/SKILL.md << 'EOF'
---
name: my-skill
description: What this skill does
---

# My Skill

Instructions for Claude...
EOF

4. Give Claude Verification Methods

The most important tip: Give Claude a way to verify its work.

• Run tests after changes
• Use browser testing for UI
• Check linter output
• Verify with git diff

5. Continuously Update CLAUDE.md

After every mistake or correction:

"Update CLAUDE.md so you don't make that mistake again"

Over time, Claude's error rate will measurably drop.

6. Use Voice Dictation

On macOS, press fn twice to dictate. You speak 3x faster than you type, and your prompts become more detailed.

Repository Structure

ChernyCode/
├── README.md
├── CLAUDE.md                      # Project memory for Claude Code
├── AGENTS.md                      # Agent instructions for Cursor
├── threads.md                     # Source material from Boris Cherny
│
├── .cursor/skills/                # Project-specific Cursor skills
│   ├── code-style/SKILL.md
│   ├── testing/SKILL.md
│   ├── git-workflow/SKILL.md
│   └── llm-development/SKILL.md
│
├── claude_personal_skills/        # Install to ~/.claude/skills/
│   ├── commit-push-pr/SKILL.md
│   ├── techdebt/SKILL.md
│   └── code-simplifier/SKILL.md
│
├── claude_subagents/              # Install to ~/.claude/agents/
│   ├── code-reviewer.md
│   ├── test-writer.md
│   └── doc-generator.md
│
├── cursor_personal_skills/        # Install to ~/.cursor/skills-cursor/
│   ├── commit-push-pr/SKILL.md
│   ├── techdebt/SKILL.md
│   └── code-simplifier/SKILL.md
│
├── cursor_subagents/              # Install to ~/.cursor/agents/
│   ├── code-reviewer.md
│   ├── doc-generator.md
│   ├── test-writer.md
│   └── verifier.md
│
└── .cursor/agents/                # Project-specific Cursor subagents
    ├── code-reviewer.md
    ├── doc-generator.md
    ├── test-writer.md
    └── verifier.md

Target Installation Structure

After installation, your home directory will have:

~/.claude/
├── CLAUDE.md              # Personal memory (all projects)
├── skills/
│   ├── commit-push-pr/SKILL.md
│   ├── techdebt/SKILL.md
│   └── code-simplifier/SKILL.md
└── agents/
    ├── code-reviewer.md
    ├── test-writer.md
    └── doc-generator.md

~/.cursor/
├── skills-cursor/
│   ├── commit-push-pr/SKILL.md
│   ├── techdebt/SKILL.md
│   └── code-simplifier/SKILL.md
└── agents/
    ├── code-reviewer.md
    ├── doc-generator.md
    ├── test-writer.md
    └── verifier.md

Customization

Edit Project Memory

Update CLAUDE.md and AGENTS.md with:

• Your project's purpose and architecture
• Coding standards specific to your team
• Common commands and workflows
• Known pitfalls (add these as you encounter them)

Create New Skills

1. Create a directory: ~/.claude/skills/my-skill/
2. Add SKILL.md with frontmatter and instructions
3. Invoke with /my-skill

Create New Subagents

For Claude Code:

1. Create a file: ~/.claude/agents/my-agent.md
2. Add frontmatter with name, description, allowed-tools
3. Add the agent's system prompt
4. Use by asking Claude to "use the my-agent agent"

For Cursor:

1. Create a file: .cursor/agents/my-agent.md (project) or ~/.cursor/agents/my-agent.md (global)
2. Add YAML frontmatter with:
   • name: Unique identifier (lowercase, hyphens)
   • description: When to use (Agent reads this for auto-delegation)
   • model: fast, inherit, or specific model ID
   • readonly: Set to true for read-only operations
   • is_background: Set to true for background execution
3. Add the agent's prompt below the frontmatter
4. Invoke with /my-agent or mention naturally in your prompt

Sources

Based on Boris Cherny's threads:

• How I use Claude Code
• Tips from the Claude Code team

License

MIT