CLAUDE.md Management
Skill1 starsChanged 7 months agoMITView on GitHub
Create and maintain CLAUDE.md project configuration files. Define project context, coding standards, tool preferences, and persistent memory for Claude Code sessions. Trigger phrases: "create CLAUDE.md", "update CLAUDE.md", "project config", "set up claude config", "configure claude for this project", "project context file"
In this repository164 other files it gives its agents.
- instrument-data-to-allotropeSkill
- literature-reviewSkill
- nextflow-developmentSkill
- scientific-problem-selectionSkill
- scvi-toolsSkill
- single-cell-rna-qcSkill
- discover-brandSkill
- enforce-voiceSkill
- generate-guidelinesSkill
- csat-reportSkill
- customer-escalationSkill
- customer-researchSkill
- draft-responseSkill
- kb-articleSkill
- sentiment-analysisSkill
- ticket-triageSkill
- analyzeSkill
- build-dashboardSkill
- create-vizSkill
- data-context-extractorSkill
- data-modelingSkill
- data-pipelineSkill
- etl-designSkill
- explore-dataSkill
- sql-queriesSkill
- statistical-analysisSkill
- validate-dataSkill
- write-querySkill
- accessibility-reviewSkill
- component-librarySkill
- design-critiqueSkill
- design-handoffSkill
- design-systemSkill
- frontend-designSkill
- research-synthesisSkill
- user-researchSkill
- ux-copySkill
- wireframingSkill
- Agent DevelopmentSkill
- MCP Server BuilderSkill
- Plugin DevelopmentSkill
- Prompt EngineeringSkill
- Skill CreatorSkill
- curriculum-designSkill
- lesson-planningSkill
- quiz-generatorSkill
- study-guideSkill
- tutoringSkill
- api-designSkill
- architectureSkill
- ci-cd-pipelineSkill
- code-reviewSkill
- database-migrationSkill
- debugSkill
- dependency-auditSkill
- deploy-checklistSkill
- documentationSkill
- error-handlingSkill
- git-workflowSkill
- incident-responseSkill
- The first 60 shown.
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 11 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈11
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 15
Create and maintain CLAUDE.md files that give Claude Code persistent context about your project, coding standards, and preferences.
- Very specific: line 29
Read package.json, pyproject.toml, Cargo.toml, or equivalent for tech stack
- Very specific: line 30
Check for existing linter configs (.eslintrc, .prettierrc, ruff.toml)
- Distinctive: line 31
Look for existing documentation (README.md, CONTRIBUTING.md)
- Distinctive: line 33
Check for CI/CD configuration files
- Distinctive: line 38
Ask the user about preferences not discoverable from the codebase:
- Distinctive: line 93
Use clear, imperative language ("Use camelCase for variables")
- Distinctive: line 94
Keep each section concise and scannable
- Distinctive: line 97
Add warnings for common pitfalls
- Distinctive: line 101
Verify the CLAUDE.md is effective:
- Distinctive: line 112
Update the CLAUDE.md when:
- Very specific: line 145
Action: Read pyproject.toml, check for ruff/mypy configs, identify the pipeline framework, and generate a CLAUDE.md with data engineering …
- Distinctive: line 161
Version control it: - commit CLAUDE.md so the whole team benefits
- Distinctive: line 162
Avoid duplication: - do not repeat what linter configs already enforce
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
CLAUDE.md Management
Create and maintain CLAUDE.md files that give Claude Code persistent context about your project, coding standards, and preferences.
Prerequisites
- A project directory to configure
- Understanding of the project's tech stack and conventions
- Knowledge of team coding standards and preferences
Workflow
Step 1: Assess the Project
Scan the project to understand its structure:
- Read
package.json,pyproject.toml,Cargo.toml, or equivalent for tech stack - Check for existing linter configs (
.eslintrc,.prettierrc,ruff.toml) - Look for existing documentation (README.md, CONTRIBUTING.md)
- Identify the test framework and test directory structure
- Check for CI/CD configuration files
- Note the directory layout and naming conventions
Step 2: Gather Preferences
Ask the user about preferences not discoverable from the codebase:
- Code style - beyond what linters enforce (naming conventions, file organization)
- Architecture patterns - MVC, hexagonal, functional, etc.
- Commit conventions - conventional commits, ticket references, etc.
- Testing philosophy - TDD, coverage targets, test naming
- Tool preferences - preferred commands for build, test, deploy
- Forbidden patterns - things Claude should never do in this project
Step 3: Structure the CLAUDE.md
Organize the file with these sections:
# Project Overview
Brief description of what the project does and its key technologies.
# Tech Stack
- Language, framework, and major dependencies
- Database and infrastructure details
# Project Structure
Key directories and their purposes.
# Coding Standards
- Naming conventions
- File organization rules
- Import ordering
- Error handling patterns
# Commands
- Build: `<command>`
- Test: `<command>`
- Lint: `<command>`
- Deploy: `<command>`
# Testing
- Framework and test file conventions
- How to run specific tests
- Coverage expectations
# Git Workflow
- Branch naming conventions
- Commit message format
- PR requirements
# Important Notes
- Critical warnings or gotchas
- Things Claude should always or never doStep 4: Write the CLAUDE.md
Generate the file using the Write tool. Place it at the project root:
- Use clear, imperative language ("Use camelCase for variables")
- Keep each section concise and scannable
- Include actual commands, not placeholders
- List concrete examples for ambiguous rules
- Add warnings for common pitfalls
Step 5: Validate the Configuration
Verify the CLAUDE.md is effective:
- Commands listed actually work when run
- Referenced directories and files exist
- Coding standards match existing code patterns
- No contradictory instructions
- File is under 200 lines for optimal context usage
- Critical rules are near the top of the file
Step 6: Maintain Over Time
Update the CLAUDE.md when:
- New dependencies or tools are added
- Coding standards evolve
- Project structure changes
- New team conventions are adopted
- Common mistakes are identified
Output Format
A single CLAUDE.md file at the project root containing:
- Project overview in 2-3 sentences
- Tech stack list
- Project structure summary
- Coding standards as bullet points
- Runnable commands for common tasks
- Testing conventions
- Git workflow rules
- Important warnings and notes
Examples
Example 1: Node.js Web Application
User: "Set up CLAUDE.md for our Express API"
Action: Scan package.json for dependencies, check for ESLint/Prettier configs, identify test framework, and generate a CLAUDE.md with API-specific conventions (route naming, middleware patterns, error response format).
Example 2: Python Data Pipeline
User: "Create CLAUDE.md for our data pipeline project"
Action: Read pyproject.toml, check for ruff/mypy configs, identify the pipeline framework, and generate a CLAUDE.md with data engineering conventions (typing requirements, logging standards, idempotency rules).
Example 3: Monorepo Setup
User: "Configure CLAUDE.md for our monorepo"
Action: Scan the workspace root and each package, identify shared tooling, and generate a root CLAUDE.md with cross-package rules plus guidance on per-package conventions.
Best Practices
- Be specific over general - "Use
zodfor validation" beats "validate inputs" - Keep it short - Claude reads this every session; respect the context window
- Put critical rules first - most important instructions at the top
- Use imperative mood - "Run tests before committing" not "Tests should be run"
- Include real commands - paste actual CLI commands, not descriptions
- Update regularly - stale CLAUDE.md causes incorrect behavior
- Version control it - commit CLAUDE.md so the whole team benefits
- Avoid duplication - do not repeat what linter configs already enforce
Error Handling
| Error | Cause | Recovery |
| Conflicting rules | Two sections contradict each other | Audit and resolve conflicts, keep one source of truth |
| Stale commands | Build/test commands changed | Re-scan project config files and update |
| Overly long file | Too much detail | Trim to essentials, move details to docs |
| Missing context | Key conventions not documented | Interview the team, add missing sections |
| Wrong tech stack | Project evolved since CLAUDE.md was written | Re-run assessment step |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

