A comprehensive configuration repository that combines ClaudeKit, custom agents/commands, and Claude Code official features for an optimal AI-assisted development workflow.
Prerequisites: Node.js 22.14+ Check version:
node --version| Install: https://nodejs.org⚠️ Recommended: Claude Code Max plan
Install claudeflow globally via your preferred package manager:
npm install -g @33strategies/claudeflowyarn global add @33strategies/claudeflowpnpm add -g @33strategies/claudeflowAfter installation, run the setup command:
claudeflow setupChoose your installation mode:
- 1) Global - Install to
~/.claude/(available in all projects) - 2) Project - Install to
./.claude/(this project only)
Quick decision:
- Solo developer or want commands everywhere? → Global mode
- Team project or want version control? → Project mode
- Both? → You can do both! They work together via configuration hierarchy.
The setup will:
- Check prerequisites (Node.js 22.14+, npm, Claude Code CLI)
- Verify ClaudeKit installation (installed automatically as dependency)
- Copy custom workflow commands to your chosen location
- Set up configuration files
- Run ClaudeKit setup
Optional but recommended: simple-task-master is already included for enhanced task tracking:
stm list --pretty --tag feature:<slug>This enables /spec:decompose and /spec:execute to automatically track tasks.
This repository implements a layered configuration approach:
- ClaudeKit Foundation (installed via npm) - 30+ agents, 20+ commands, 25+ hooks
- Custom Extensions (this repo) - Domain-specific agents and workflow commands
- Official Claude Code Features - Built-in capabilities and plugin system
All three layers work together seamlessly in Claude Code.
- 30+ specialized agents (TypeScript, React, Testing, Database, DevOps, etc.)
- 20+ workflow commands (/git:commit, /spec:create, /research, etc.)
- 25+ intelligent hooks (file-guard, linting, testing, checkpoints)
- Automated quality assurance and error prevention
- 4 custom workflow commands:
- /ideate: Structured ideation with comprehensive documentation
- /ideate-to-spec: Transform ideation into validated specification
- /spec:feedback: Post-implementation feedback with interactive decisions
- /spec:doc-update: Parallel documentation review based on specs
- 4 enhanced spec command overrides (replace ClaudeKit versions):
- /spec:create: Detects output path and creates specs in feature directories
- /spec:decompose: Incremental mode preserves completed work, tags STM tasks
- /spec:execute: Session resume capability with implementation tracking
- /spec:migrate: Migrates existing specs to feature-directory structure
- Complete end-to-end workflow from ideation to deployment
- Example configurations for teams and individuals
- Uses ClaudeKit's 30+ agents for specialized tasks
- Task tracking: Integrates with simple-task-master (stm) - use
stm list --pretty --tag feature:<slug>to track progress
- 5-tier configuration hierarchy
- Project-level
.claude/directories - CLAUDE.md context files
- MCP server integration
- Plugin system
Check that everything is set up correctly:
claudeflow doctorThis diagnostic command checks:
- Node.js version (need 22.14+)
- npm availability
- Claude Code CLI installation
- ClaudeKit installation
- .claude/ directory structure
- Command files presence
- Ensure npm global bin is in your PATH
- Try:
npm list -g @33strategies/claudeflow - Reinstall:
npm install -g @33strategies/claudeflow
- ClaudeKit should install automatically as a dependency
- Manual install:
npm install -g claudekit - Verify:
claudekit --version
- Run:
claudeflow doctor - Verify files exist in
~/.claude/commands/(global) or./.claude/commands/(project) - Restart Claude Code
Run the diagnostic command for detailed information:
claudeflow doctorIf you previously used the bash script installation:
-
Uninstall old version:
# Global installation rm -rf ~/.claude # OR Project installation rm -rf .claude
-
Install via npm:
npm install -g @33strategies/claudeflow
-
Run setup:
# Global (if you used install.sh user) claudeflow setup --global # OR Project (if you used install.sh project) claudeflow setup --project
-
Verify installation:
claudeflow doctor
Note: Your existing STM tasks will be preserved. The migration only affects the installation method, not your feature work or task data.
claudeflow is published with npm provenance attestations and SLSA Level 2 compliance. All releases are built and published via GitHub Actions with cryptographic verification.
For details on verifying package authenticity, supply chain security, and security best practices, see docs/SECURITY.md.
claudeflow checks for updates automatically once per week (every 7 days). This check runs in the background and doesn't block command execution.
Manual update check:
npm outdated -g @33strategies/claudeflowWhen updates are available, you'll see a notification:
╭───────────────────────────────────────────────────╮
│ │
│ Update available: 1.3.0 │
│ Current version: 1.2.0 │
│ Run: npm install -g @33strategies/claudeflow │
│ │
╰───────────────────────────────────────────────────╯
To update:
# npm
npm update -g @33strategies/claudeflow
# yarn
yarn global upgrade @33strategies/claudeflow
# pnpm
pnpm update -g @33strategies/claudeflowAll documents related to a feature are organized in a single directory:
specs/
└── <feature-slug>/
├── 01-ideation.md # Ideation and research
├── 02-specification.md # Validated specification
├── 03-tasks.md # Task breakdown
├── 04-implementation.md # Implementation summary
└── 05-feedback.md # Post-implementation feedback log
Benefits:
- All related documents in one place
- Clear lifecycle progression (01 → 02 → 03 → 04)
- Easy to find and navigate
- Git-friendly tracking
- STM tasks tagged with
feature:<slug>for filtering
Example:
specs/add-user-auth-jwt/
├── 01-ideation.md # Created by /ideate
├── 02-specification.md # Created by /ideate-to-spec → /spec:create
├── 03-tasks.md # Created by /spec:decompose
├── 04-implementation.md # Created by /spec:execute
└── 05-feedback.md # Created by /spec:feedback (after testing)
claudeflow/ (@33strategies/claudeflow npm package)
├── package.json # npm package metadata
├── bin/
│ └── claudeflow.js # CLI entry point
├── lib/
│ ├── setup.js # Installation logic
│ └── doctor.js # Diagnostic command
├── scripts/
│ └── verify-files.js # Pre-publish file verification
│
├── .claude/ # Custom configuration (distributed in package)
│ ├── commands/ # Custom slash commands
│ │ ├── ideate.md # Structured ideation workflow
│ │ ├── ideate-to-spec.md # Transform ideation to spec
│ │ └── spec/
│ │ ├── create.md # Enhanced spec creation
│ │ ├── decompose.md # Incremental task breakdown
│ │ ├── execute.md # Session-aware implementation
│ │ ├── feedback.md # Post-implementation feedback
│ │ ├── doc-update.md # Documentation review
│ │ └── migrate.md # Spec structure migration
│ ├── settings.json.example # Example configuration
│ └── README.md # Component documentation
│
├── .claude-plugin/ # Plugin metadata
│ └── plugin.json # Package metadata
│
├── templates/ # Configuration templates
│ ├── project-config/ # Project-level templates
│ │ ├── settings.json # Team settings
│ │ ├── settings.local.json.example
│ │ ├── CLAUDE.md # Project instructions
│ │ ├── CLAUDE.local.md.example
│ │ └── .gitignore.example
│ └── user-config/ # User-level templates
│ ├── settings.json # Personal global settings
│ └── CLAUDE.md # Personal preferences
│
├── docs/ # Documentation
│ ├── DESIGN_RATIONALE.md # Design validation & best practices
│ └── ...
│
├── .github/
│ └── workflows/
│ └── release.yml # CI/CD automation
├── .releaserc.json # Semantic-release config
├── .npmignore # Package exclusion rules
├── LICENSE # MIT License
├── CHANGELOG.md # Version history (auto-generated)
└── README.md # This file
┌─────────────────────────────────────┐
│ Layer 3: Custom (This Repo) │ ← Domain-specific agents & commands
├─────────────────────────────────────┤
│ Layer 2: ClaudeKit (npm) │ ← 30+ agents, 20+ commands, 25+ hooks
├─────────────────────────────────────┤
│ Layer 1: Claude Code (Official) │ ← Base CLI, plugin system, MCP
└─────────────────────────────────────┘
| Feature | Claude Code | ClaudeKit | This Repo |
|---|---|---|---|
| Agents | Built-in agents | typescript-expert, react-expert, database-expert, etc. (30+) | Uses ClaudeKit agents |
| Commands | /init, /plugin |
/git:commit, /spec:create, /research, etc. (20+) | /ideate, /ideate-to-spec, /spec:doc-update + overrides for /spec:create, /spec:decompose, /spec:execute, /spec:migrate |
| Hooks | Hook system | file-guard, typecheck-changed, lint-changed, etc. (25+) | Uses ClaudeKit hooks via settings.json |
| Config | .claude/ hierarchy | Integrates with .claude/ | Custom commands in .claude/ |
Claude Code uses a 5-tier precedence system:
- Enterprise policies (highest) - IT-managed, cannot be overridden
- CLI arguments - Temporary session overrides
- Local settings (
.claude/settings.local.json) - Personal, gitignored - Project settings (
.claude/settings.json) - Team, committed - User settings (
~/.claude/settings.json) - Personal global defaults
Structured ideation workflow that enforces complete investigation for any code-change task (bug fix or feature). Produces comprehensive ideation documentation including:
- Intent & assumptions
- Pre-reading log
- Codebase mapping
- Root cause analysis (for bugs)
- Research findings
- Clarification questions
Usage: /ideate Fix chat UI auto-scroll bug when messages exceed viewport height
Transform an ideation document into a validated, implementation-ready specification. This command:
- Reads and synthesizes the ideation document
- Interactively gathers decisions from the user
- Creates a detailed specification using
/spec:create - Detects and resolves open questions interactively
- Validates it with
/spec:validate - Presents a summary with next steps
Interactive Question Resolution: After creating the specification, the system automatically:
- Detects unanswered questions in the "Open Questions" section
- Presents each question with context and options
- Records answers in strikethrough format (audit trail)
- Re-validates and loops until all questions resolved
- Shows progress: "Question 3 of 12"
- Supports multi-select for questions requiring multiple choices
- Detects external edits to prevent data loss
- Backward compatible (skips if no questions exist)
- Re-entrant (skips already-answered questions)
- Save-as-you-go for recoverability
Usage: /ideate-to-spec specs/<slug>/01-ideation.md
Example with Questions:
/ideate-to-spec specs/my-feature/01-ideation.md
# → Creates specification
# → System detects 5 open questions
# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
# Question 1 of 5
# ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
# Package Manager Support
# → User selects: npm, yarn, pnpm (all three)
# → Spec updated with strikethrough answer
# → Continues with questions 2-5
# → Final summary shows 5 questions resolvedAnswer Format (Audit Trail):
Questions are marked as resolved using strikethrough format:
Before:
1. **ClaudeKit Version Compatibility**
- Option A: Pin exact version
- Option B: Use caret rangeAfter:
1. ~~**ClaudeKit Version Compatibility**~~ (RESOLVED)
**Answer:** Use caret range (^1.0.0)
**Rationale:** Automatic updates, test compatibility in CI/CD
Original context preserved:
- Option A: Pin exact version
- Option B: Use caret rangeThis format provides a complete audit trail showing both the original question and the final decision.
Process ONE specific piece of post-implementation feedback from testing or usage. This command:
- Validates prerequisites (implementation must exist)
- Prompts for detailed feedback description
- Explores relevant code with targeted investigation
- Optionally consults research-expert for solution approaches
- Guides through interactive decisions (implement/defer/out-of-scope)
- Updates spec changelog for "implement now" decisions
- Creates STM tasks for deferred feedback
- Logs all decisions in
05-feedback.md
Integrates with incremental /spec:decompose and resume /spec:execute for seamless iteration.
Usage: /spec:feedback specs/my-feature/02-specification.md
Review all documentation to identify what needs to be updated based on a new specification file. Launches parallel documentation expert agents to review each doc file for:
- Deprecated content
- Content requiring updates
- Missing content for new features
Usage: /spec:doc-update specs/text-generator-spec.md
These commands override ClaudeKit's versions with enhanced feature-directory organization and STM integration:
Enhanced version that detects output paths and organizes specs in feature directories. Creates specs in specs/<slug>/02-specification.md format.
Usage: /spec:create Add user authentication with JWT tokens
Enhanced version with incremental mode that detects changelog updates and creates only new tasks while preserving completed work. Extracts feature slugs and tags all STM tasks with feature:<slug> for filtering. Creates task breakdown in specs/<slug>/03-tasks.md with re-decompose metadata.
Usage: /spec:decompose specs/add-user-auth/02-specification.md
Incremental Mode: Automatically detects when spec changelog has been updated after feedback and creates only new tasks for changes.
Enhanced version with session resume capability that continues from previous progress. Reads implementation summary to skip completed tasks, resume in-progress work, and append new session history. Creates/updates specs/<slug>/04-implementation.md with cross-session context.
Usage: /spec:execute specs/add-user-auth/02-specification.md
Resume Mode: Automatically detects previous sessions and skips completed work, maintaining full implementation history.
Session Continuity: When you run /spec:execute multiple times on the same spec, it reads 04-implementation.md to understand what work has already been completed. This allows you to:
- Work on large features across multiple sessions without re-doing completed tasks
- Return to implementation after testing/feedback cycles
- Maintain a complete history of all sessions (Session 1, Session 2, etc.)
- Seamlessly integrate feedback workflow: implement → test →
/spec:feedback→/spec:decompose(incremental) →/spec:execute(resume)
Track Progress: stm list --pretty --tag feature:add-user-auth (replaces removed /spec:progress)
Migrates existing specs from flat structure to feature-directory structure. Moves specs, tasks, and ideation docs into organized specs/<slug>/ directories and tags STM tasks.
Usage: /spec:migrate
Build Tools: webpack-expert, vite-expert
Languages: typescript-expert, typescript-build-expert, typescript-type-expert
Frontend: react-expert, react-performance-expert, nextjs-expert, css-styling-expert, accessibility-expert
Testing: testing-expert, jest-testing-expert, vitest-testing-expert, playwright-expert
Database: database-expert, postgres-expert, mongodb-expert
DevOps: docker-expert, github-actions-expert, devops-expert, git-expert
Specialized: ai-sdk-expert, nestjs-expert, kafka-expert, loopback-expert, nodejs-expert
Utilities: code-search, linting-expert, refactoring-expert, code-review-expert, research-expert
Git Workflow: /git:status, /git:commit, /git:checkout, /git:push, /git:ignore-init
Specifications: /spec:validate
- Note: This repo overrides
/spec:create,/spec:decompose,/spec:executewith enhanced versions - Enhanced versions add feature-directory organization and STM task tagging
/spec:migrateis a custom addition for migrating existing specs- Use
stm list --pretty --tag feature:<slug>to track progress (replaces removed/spec:progress)
Quality: /code-review, /validate-and-fix
Research: /research
Development: /create-command, /create-subagent
Checkpoints: /checkpoint:restore
Hooks: /hook:disable, /hook:enable, /hook:status
PreToolUse: file-guard (protects sensitive files)
PostToolUse: typecheck-changed, lint-changed, check-any-changed, test-changed, codebase-map-update, check-comment-replacement
Stop: create-checkpoint, check-todos, typecheck-project, lint-project, test-project, self-review
UserPromptSubmit: thinking-level, codebase-map
This repository implements a complete end-to-end workflow for feature development:
┌─────────────────────────────────────────────────────────────────┐
│ IDEATION PHASE │
└─────────────────────────────────────────────────────────────────┘
│
▼
/ideate <task-brief>
(Custom Command)
│
Creates ideation doc
with research & analysis
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ SPECIFICATION PHASE │
└─────────────────────────────────────────────────────────────────┘
│
▼
/ideate-to-spec <ideation-doc>
(Custom, calls /spec:create & /spec:validate)
│
Creates validated specification
│
▼
/spec:decompose <spec-file>
(Custom Override - uses stm if installed)
│
Breaks spec into tasks
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ IMPLEMENTATION PHASE │
└─────────────────────────────────────────────────────────────────┘
│
▼
/spec:execute <spec-file>
(Custom Override - uses stm if installed)
│
Implements tasks
│
▼
stm list --pretty --tag feature:<slug>
(Track progress)
│
View task status and completion
│
┌──────────┴──────────┐
│ │
Not Finished Finished
│ │
└──────────┬──────────┘
│
▼
Manual Testing
│
Discover issues or
improvement opportunities
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ FEEDBACK PHASE │
└─────────────────────────────────────────────────────────────────┘
│
▼
/spec:feedback <spec-file>
(Custom Command - one item at a time)
│
Process feedback item:
• Code exploration
• Optional research
• Interactive decisions
│
┌──────────┴──────────┬────────────┐
│ │ │
Implement Now Defer Out of Scope
│ │ │
Update spec changelog Create STM task Log only
│ │ │
▼ │ │
/spec:decompose (incremental) │ │
│ │ │
/spec:execute (resume) │ │
│ │ │
└──────────┬──────────┴────────────┘
│
More feedback items? (repeat)
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ COMPLETION PHASE │
└─────────────────────────────────────────────────────────────────┘
│
▼
/git:commit
(ClaudeKit Command)
│
Commits implementation
│
▼
/spec:doc-update <spec-file>
(Custom Command)
│
Reviews & updates documentation
│
▼
/git:commit
(ClaudeKit Command)
│
Commits doc updates
│
▼
/git:push
(ClaudeKit Command)
│
Pushes to remote
│
▼
DONE! 🎉
- Ideation → Comprehensive investigation and research
- Specification → Validated, implementation-ready spec with interactive question resolution
- Decomposition → Tasks broken down with dependencies (uses stm if installed, tags with
feature:<slug>) - Implementation → Iterative execution with stm task tracking via
stm list --pretty --tag feature:<slug> - Feedback → Process post-implementation feedback with structured decisions (implement/defer/out-of-scope)
- Completion → Documentation updates and git workflow
Note: Phase 2 (Specification) now includes automatic open questions resolution to ensure specs are implementation-ready before decomposition begins.
# Step 1: Start with ideation
/ideate Add user authentication with JWT tokens
# → Creates: specs/add-user-auth-jwt/01-ideation.md
# Includes: investigation, research, clarifications
# Step 2: Transform to validated specification
/ideate-to-spec specs/add-user-auth-jwt/01-ideation.md
# → User makes decisions interactively (Step 2: ideation clarifications)
# → System creates specification via /spec:create
# → System detects and resolves open questions (Step 6: interactive resolution)
# → All questions answered interactively with progress tracking
# → Specification re-validated until complete
# → Summary includes both ideation decisions AND resolved spec questions
# → Creates: specs/add-user-auth-jwt/02-specification.md (validated & complete)
# Step 3: Break down into tasks
/spec:decompose specs/add-user-auth-jwt/02-specification.md
# → Creates: specs/add-user-auth-jwt/03-tasks.md
# → Creates STM tasks tagged with feature:add-user-auth-jwt
# Step 4: Start implementation
/spec:execute specs/add-user-auth-jwt/02-specification.md
# → Implements tasks incrementally, updating stm status
# Step 5: Check progress (can run anytime during implementation)
stm list --pretty --tag feature:add-user-auth-jwt
# → Shows: completion %, current phase, task status
# Note: This replaces the removed /spec:progress command
# Step 6: Continue implementing (loop back to step 4 if needed)
/spec:execute specs/add-user-auth-jwt/02-specification.md
# Step 7: Final progress check (should show 100%)
stm list --pretty --tag feature:add-user-auth-jwt
# Step 8: Implementation summary created automatically
# → Creates: specs/add-user-auth-jwt/04-implementation.md
# Step 9: Manual testing (discover feedback items)
# Test the implemented feature manually
# Identify bugs, improvements, or missing functionality
# Step 10: Process feedback (repeat for each feedback item)
/spec:feedback specs/add-user-auth-jwt/02-specification.md
# → Interactive workflow:
# 1. Describe feedback item
# 2. Code exploration
# 3. Optional research
# 4. Make decision:
# - Implement Now: Updates spec changelog
# - Defer: Creates STM task for later
# - Out of Scope: Logs decision only
# Step 11: If "Implement Now" was chosen, run incremental decompose
/spec:decompose specs/add-user-auth-jwt/02-specification.md
# → Incremental mode: preserves completed tasks, creates only new ones
# Step 12: Resume implementation for new tasks
/spec:execute specs/add-user-auth-jwt/02-specification.md
# → Resume mode: skips completed work, implements new feedback tasks
# Step 13: Repeat steps 9-12 until all feedback is addressed
# Step 14: Commit implementation
/git:commit
# → Creates conventional commit with changes
# Step 15: Update documentation
/spec:doc-update specs/add-user-auth-jwt/02-specification.md
# → Parallel agents review all docs
# → Identifies outdated content and missing docs
# Step 16: Commit documentation updates
/git:commit
# Step 17: Push to remote
/git:push
# All documents for this feature are now in: specs/add-user-auth-jwt/
# ├── 01-ideation.md
# ├── 02-specification.md
# ├── 03-tasks.md
# ├── 04-implementation.md
# └── 05-feedback.mdIf you already know what you need:
# Start directly with spec creation
/spec:create Add user authentication with JWT tokens
# Then follow steps 3-11 aboveIf you have specs in the old flat structure:
# Migrate all existing specs to new structure
/spec:migrate
# This will:
# - Move specs/*.md to specs/<slug>/02-specification.md
# - Move specs/*-tasks.md to specs/<slug>/03-tasks.md
# - Move docs/ideation/*.md to specs/<slug>/01-ideation.md
# - Tag STM tasks with feature:<slug>
# - Generate migration report# After implementing a feature via a spec
/spec:doc-update specs/add-user-auth-jwt/02-specification.md
# This will:
# - Launch parallel documentation expert agents
# - Review all docs for outdated content
# - Identify missing documentation for new features
# - Provide prioritized recommendationscd your-project
# Initialize with ClaudeKit + custom config
./path/to/claude-config/install.sh project
# Customize settings
cp .claude/settings.json.example .claude/settings.json
# Edit .claude/settings.json for team
# Add to git
echo ".claude/settings.local.json" >> .gitignore
echo "CLAUDE.local.md" >> .gitignore
git add .claude/ CLAUDE.md .gitignore
git commit -m "Add Claude Code configuration"# Install to ~/.claude
./path/to/claude-config/install.sh user
# Customize personal settings
cp ~/.claude/settings.json ~/.claude/settings.local.json
# Edit ~/.claude/settings.local.json for personal preferencesTeam Settings (.claude/settings.json - committed):
{
"permissions": {
"allow": ["Read", "Edit", "Bash(git:*)"],
"deny": [".env", "**/*.key", "secrets/"]
},
"hooks": {
"PreToolUse": [{
"matcher": "Read|Edit|Write",
"hooks": [{"type": "command", "command": "claudekit-hooks run file-guard"}]
}]
}
}Personal Overrides (.claude/settings.local.json - gitignored):
{
"environmentVariables": {
"EDITOR": "code"
}
}# Check for updates
npm outdated -g | grep claudekit
# Update to latest
npm update -g claudekit# Pull latest from this repo
cd path/to/claude-config
git pull origin main
# Re-run installation to update
./install.sh project# Validate agents
claudekit lint-agents
# Validate commands
claudekit lint-commands
# List all components
claudekit list agents
claudekit list commands
# Check hook status
claudekit-hooks status
# Profile hook performance
claudekit-hooks profile- Always enable file-guard hook
- Never commit secrets to
.claude/settings.json - Use environment variable references:
"API_KEY": "${API_KEY}" - Keep
.envfiles gitignored
- Commit
.claude/settings.jsonandCLAUDE.mdto git - Gitignore
.claude/settings.local.jsonandCLAUDE.local.md - Document custom agents and commands in README
- Use PR reviews for configuration changes
- Profile hooks regularly:
claudekit-hooks profile - Keep UserPromptSubmit hooks under 10k characters
- Use
*-changedhooks for fast feedback - Use
*-projecthooks sparingly (slow but comprehensive)
- Start with 10-15 most relevant agents
- Add more as needed
- Remove unused agents to reduce context overhead
- Document agent purposes in
.claude/agents/README.md
- Keep CLAUDE.md focused and concise
- Use file references instead of embedding code
- Enable codebase-map hook for project structure
- Break large tasks into smaller steps
# Reinstall globally
npm install -g claudekit
# Check installation
which claudekit
claudekit --version# Verify agent files
ls -la .claude/agents/
# Validate agents
claudekit lint-agents
# Restart Claude Code session# Check hook status
claudekit-hooks status
# Enable hooks
claudekit-hooks enable file-guard
# Verify settings.json syntax
cat .claude/settings.json | jq .
# Restart Claude Code session- Check configuration hierarchy (local > project > user)
- Restart Claude Code session
- Validate JSON syntax:
jq . .claude/settings.json - Check for enterprise policies overriding settings
- Installation Guide - When to use user vs project installation
- Setup Guide - Detailed setup instructions
- Design Rationale - Design validation and best practices
- Feedback Workflow Guide - Complete guide to post-implementation feedback
- GitHub Issues - Report problems
Contributions welcome! Please:
- Follow existing patterns and structure
- Test changes thoroughly
- Update documentation
- Submit PRs with clear descriptions
MIT License - feel free to use and modify for your projects.
- ClaudeKit by Carl Rannaberg - Comprehensive toolkit foundation
- Anthropic - Claude Code CLI and official features
- Community contributors and examples
Note: This repository implements a validated, research-backed approach for managing Claude Code configuration. See research.md for detailed analysis, sources, and best practices.
- ✅ Simpler: npm install vs. submodule complexity
- ✅ Standard: Familiar to all Node.js developers
- ✅ Easy updates:
npm update -gvs.git submodule update --remote - ✅ No merge conflicts with upstream
- ❌ Submodules add unnecessary complexity
- ✅ Automated setup via installation script
- ✅ Validated configurations from research
- ✅ Combines best of ClaudeKit + custom + official
- ✅ Easy to maintain and update
- ✅ More flexible than plugins alone
- ✅ Can customize ClaudeKit features
- ✅ Works with npm-distributed tools
- ✅ Better for teams with specific needs
This hybrid approach provides the best of all worlds: ClaudeKit's mature foundation, your custom extensions, and official Claude Code features, all working together seamlessly.
