Identity & Working Style

Tell Claude who you are and how you want to be treated. This shapes every interaction.

Your Name Claude will use this name instead of generic terms like "user."

Your Role

Team Context

Solo developer I'm the only one touching this codebase. Small team (2-5) A few collaborators. Conventions matter. Larger team (6+) Multiple contributors. Standards are critical.

Should Claude push back when you're wrong?

Yes Recommended Question my assumptions. Accuracy over agreeableness. Gently Suggest alternatives without direct pushback. No Just do what I ask.

Communication Style

Direct and concise No fluff. Get to the point. Bullet points over walls of text. Detailed explanations Explain the why, not just the what. Context matters. Balanced Concise by default, detailed when it matters.

Project Context

Help Claude understand what it's working with.

Project Name

Project Type

Primary Language

Framework

Package Manager

Deploy Method

Coding Standards

Define the rules Claude should enforce in every file it touches.

Code Style Preference

Functional Pure functions, immutability, composition. Object-Oriented Classes, inheritance, encapsulation. Mixed / Pragmatic Use whatever fits the problem.

Error Handling

Fail fast Crash early, crash loudly. Don't mask errors. Graceful degradation Recover when possible, degrade when not. Defensive Validate everything. Trust nothing.

Testing Approach

TDD Write tests first. Code follows. Test after Build it, then test it. Cover happy path and edge cases. Minimal testing Critical paths only. Ship fast.

Security Standards

Enforce security standards Recommended Input sanitization, rate limiting, security headers, secrets management. Basic security only Don't commit secrets. Basic input validation. That's it. Skip for now Internal tool or prototype. I'll add this later.

Build Command (optional)

Test Command (optional)

AI Behavior

Control how much freedom Claude has and how it operates in your project.

Autonomy Level

Ask before changes Confirm every file edit, commit, and command. Ask for risky only Recommended Work freely in safe paths. Confirm deletes, force-pushes, deployments. Full autonomy Make changes freely. Trust the AI.

Commit Style

Conventional Commits feat:, fix:, chore:, docs: prefixes. Descriptive Plain English. Focus on what changed and why. Short Brief and to the point. One line.

File Creation Policy

Prefer editing existing files Only create new files when truly necessary. Create new when needed Don't hesitate to add files if the architecture calls for it.

Documentation Style

Full JSDoc / docstrings Document every public function and type. Comment complex logic only Code should be self-documenting. Comments explain why, not what. Minimal comments Let the code speak for itself.

Advanced

Optional but powerful. These sections separate a good CLAUDE.md from a great one.

Project Planning Standards

Full planning workflow Recommended Plan first, build phase by phase, test each phase, document results including failures. Lightweight Outline the plan, then build. Document the result. Skip I'll manage planning myself.

Writing Register (how Claude writes prose and docs)

Technical Professional Precise, authoritative. Steps are steps. Facts are facts. Technical but Approachable Knowledgeable but human. The reader is a peer. Casual First-person, warmer. The engineering brain is there but not running the show.

Trusted Paths (optional, comma-separated) Directories where Claude can work autonomously without asking.

Do you also use Gemini CLI?

Yes Include a Dual-AI Workflow section for Claude + Gemini coordination. No Claude Code only.

Custom Rules (optional) These will appear in a "Custom Rules" section at the end of your CLAUDE.md.

CLAUDE.md Generator

Identity & Working Style

Project Context

Coding Standards

AI Behavior

Advanced

Your CLAUDE.md is ready

Get Claude Code tips

This is 1 of 15 files in the Claude Context Kit