This repository IS the .claude configuration directory for Effect TypeScript projects following functional programming principles.

Clone this repository directly as the .claude directory in your Effect TypeScript projects:

# Clone directly into your project as .claude
cd /path/to/your/project
git clone <repo-url> .claude

# Or use as a git submodule
git submodule add <repo-url> .claude

# Or copy if you prefer not to use git
cp -r /path/to/claude-setup /path/to/your/project/.claude

After cloning, your project structure will be:

your-project/
├── .claude/           # This repository
│   ├── agents/
│   ├── hooks/
│   ├── skills/
│   └── settings.json
├── src/               # Your code
└── package.json

This repository contains:

claude-setup/              (clones into your-project/.claude/)
├── agents/               # Specialized subagents
│   ├── domain-modeler.md       # ADT domain modeling with MCP
│   ├── effect-expert.md        # Services & layers with MCP
│   ├── spec-writer.md          # Spec-driven development
│   ├── react-expert.md         # Compositional React patterns
│   └── test-writer.md          # Effect testing patterns
├── skills/               # Reusable capabilities
│   ├── typeclass-design/       # Typeclass implementation patterns
│   ├── domain-predicates/      # Predicate and Order generation
│   ├── service-implementation/ # Service design patterns
│   ├── layer-design/           # Layer composition patterns
│   ├── context-witness/        # Witness vs Capability patterns
│   └── atom-state/             # Effect Atom patterns
├── commands/             # Custom slash commands
│   ├── mailboxes.md            # Agent mailbox management
│   ├── request.md              # Inter-agent messaging
│   ├── await-mailbox.md        # Message waiting
│   └── parallelize.md          # Parallel agent execution
├── coordination/         # Multi-agent coordination
│   └── mailboxes/              # Agent communication
├── hooks/                # Hook implementations
│   ├── agent-init.sh/ts        # Session startup context injection
│   ├── skill-suggester.sh/ts   # Dynamic skill suggestions
│   └── stop-await-mailbox.sh/ts # Mailbox coordination
├── scripts/              # Utility scripts
│   ├── request.sh              # Send inter-agent messages
│   ├── await-mailbox.sh        # Wait for messages
│   ├── mailboxes.sh            # List all mailboxes
│   └── close-mailbox.sh        # Close mailboxes
├── docs/                 # Reference documentation
│   ├── project-guide.md        # Project instructions
│   └── clean-code-guide.md     # Complete patterns guide
└── settings.json         # Claude Code configuration

Claude automatically invokes the appropriate agent based on your task:

Creates type-safe domain models with:

• ADT unions for valid states
• Branded/nominal types
• Comprehensive predicates
• Order instances for sorting
• Typeclass integration
• Effect Docs: View the git subtree in .context/effect/ for best practices

Implements Effect infrastructure:

• Fine-grained service capabilities
• Layer composition
• Dependency injection
• Error handling
• Effect Docs: View the git subtree in .context/effect/ for patterns

Manages spec-driven workflow:

• Instructions → Requirements → Design → Plan
• Requires approval at each phase
• Prevents premature implementation

Compositional React patterns:

• Component composition over configuration
• No boolean props
• Effect Atom integration
• State lifting patterns

Testing strategies:

• @effect/vitest for Effect code
• Regular vitest for pure functions
• Service testing with layers

Skills extend agents with focused capabilities:

• typeclass-design: Curried signatures, dual APIs
• domain-predicates: Complete predicate/order sets
• service-implementation: Capability-based services
• layer-design: Layer composition patterns
• context-witness: Witness vs Capability decisions
• atom-state: Effect Atom reactive state

Both domain-modeler and effect-expert agents have access to the Effect documentation via the git subtree in .context/effect/.

This ensures implementations follow official Effect best practices.

Full guides available in docs/:

• docs/project-guide.md - Original project instructions
• docs/clean-code-guide.md - Complete patterns and best practices

Agents access these automatically when needed.

Automatic quality checks configured in settings.json:

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Edit|Write",
        "hooks": [
          {
            "type": "command",
            "command": "jq -r '.tool_input.file_path' | { read file_path; if echo \"$file_path\" | grep -qE '\\.(ts|tsx)$'; then bun run format && bun run lint && bun run typecheck; fi; }",
            "description": "Format, lint, and typecheck TypeScript files immediately after editing"
          }
        ]
      }
    ]
  }
}

How it works:

• PostToolUse event: Triggers immediately after Edit/Write operations
• File path extraction: Uses jq to get the modified file path from tool input
• TypeScript filter: Only runs on .ts and .tsx files
• Quality checks: Runs format → lint → typecheck in sequence
• Provides immediate feedback on code quality issues

1. Ask Claude to perform a task
2. Claude automatically:
   • Selects appropriate agent
   • Loads relevant skills
   • Accesses MCP docs if needed
   • Uses minimal context (2-10KB vs 104KB)
3. Complete task efficiently with focused expertise

After every file change:

bun run format
bun run typecheck

This is enforced in agent workflows.

"Create an Appointment domain model with status, meeting type, date, and duration"

→ Invokes domain-modeler agent (6-8KB context) → Uses domain-predicates skill → References Effect docs in .context/effect/ → Generates complete module with predicates and orders

"Implement a PaymentGateway service with handoff capability"

→ Invokes effect-expert agent (7-9KB context) → Uses service-implementation and layer-design skills → References Effect docs in .context/effect/ → Creates focused capability with proper layer

"Create a message composer component"

→ Invokes react-expert agent (7-9KB context) → Uses atom-state skill → Generates compositional structure with Effect Atom integration

• Specialized expertise per domain
• Effect documentation access via git subtree in .context/effect/ for accuracy
• Consistent patterns across projects
• Reusable configuration for all projects
• Automatic agent selection based on task

This configuration is designed for personal/team use. Adapt as needed for your projects.

To improve this configuration:

1. Test patterns in real projects
2. Update agents/skills based on learnings
3. Keep documentation in sync with code

Context pollution solved. Specialized agents ready. Effect best practices enforced. 🎉