A curated collection of specialist agents, commands, and user personas for Claude Code that enables orchestrated, user-centered development.
- a11y-expert - Accessibility specialist for inclusive web applications
- fullstack-architect - End-to-end application architecture and design
- review - Expert code reviewer for pull request analysis and feedback
Orchestrators & Planning:
/engineer- Analyze tasks and route to specialist agents/new-task- Break down complex features into actionable steps/user-story- Create user stories with persona context/misc:feature-plan- Plan features with technical specifications
React Generators:
/react:component-new- Create React components with TypeScript/react:page-new- Create Next.js pages with App Router
Skills & Configuration:
/misc:skill-builder- Build new skills with guided questions
Pre-built user personas for understanding your audience:
- End User - Non-technical users focused on task completion
- Admin - Power users managing systems and workflows
- Developer - Technical users building and integrating
Example hook scripts that run automatically at specific events:
- auto-format.sh - Auto-format code after edits (PostToolUse)
- inject-persona.sh - Inject active persona into prompts (UserPromptSubmit)
- lint-check.sh - Run linters before commits (PreToolUse)
See .claude/hooks/README.md for configuration and custom hook development.
Example project configurations showing how to set up:
- Project Context - Document tech stack, conventions, and common tasks
- MCP Servers - Configure Model Context Protocol servers per-project
git clone https://github.com/yourusername/claude-ensemble.git ~/dev/claude-ensembleOption A: Link Everything
cd your-project
ln -sf ~/dev/claude-ensemble/.claude .claudeOption B: Link Selectively
cd your-project
mkdir -p .claude
ln -sf ~/dev/claude-ensemble/.claude/agents .claude/agents
ln -sf ~/dev/claude-ensemble/.claude/commands .claude/commands
ln -sf ~/dev/claude-ensemble/.claude/contexts .claude/contextsRestart Claude Code to pick up the new configuration.
The /engineer command analyzes your request and recommends specialist agents:
/engineer I need to build a user profile page with accessibility features
Claude will analyze the task, identify required specialists (ui-architect, a11y-expert), and coordinate the work.
Create user stories that understand your audience:
/user-story As an end user, I want to filter search results
Generate React components following best practices:
/react:component-new UserProfile
Enable automated workflows with hook scripts. Example configuration in .claude/projects/{project}/hooks.json:
{
"hooks": {
"postToolUse": {
"command": ".claude/hooks/auto-format.sh",
"description": "Auto-format files after edits",
"matchers": [{
"toolName": "Edit",
"paramMatchers": {
"file_path": ".*\\.(ts|tsx|js|jsx)$"
}
}]
}
}
}See .claude/hooks/README.md for configuration options and custom hook development.
For project-specific settings like MCP servers or context, see the example project showing:
context.md- Document your project's tech stack, conventions, and tasksmcp.json- Configure MCP servers specific to your project
See .claude/projects/README.md for full details on creating and syncing project configs.
claude-ensemble/
├── .claude/
│ ├── agents/ # Deep specialist agents
│ │ ├── a11y-expert.md
│ │ ├── fullstack-architect.md
│ │ └── review.md
│ ├── commands/ # Quick utilities
│ │ ├── engineer.md # Orchestrator
│ │ ├── new-task.md
│ │ ├── user-story.md
│ │ ├── misc/
│ │ │ ├── feature-plan.md
│ │ │ └── skill-builder.md
│ │ └── react/
│ │ ├── component-new.md
│ │ └── page-new.md
│ ├── contexts/
│ │ └── personas/ # User personas
│ │ ├── admin.md
│ │ ├── developer.md
│ │ └── end-user.md
│ ├── hooks/ # Automation scripts
│ │ ├── README.md
│ │ ├── auto-format.sh
│ │ ├── inject-persona.sh
│ │ └── lint-check.sh
│ └── projects/ # Project-specific configs
│ ├── README.md
│ └── example-project/
│ ├── context.md
│ └── mcp.json
└── README.md
Orchestrated Development: Use the /engineer command to coordinate multiple specialists rather than doing everything yourself.
User-Centered: Start with personas and user stories to ensure you're building the right thing.
Best Practices: All generators follow modern framework patterns and accessibility standards.
Composable: Mix and match agents, commands, and personas for your workflow.
Contributions welcome! To add a new agent or command:
- Fork the repository
- Add your
.mdfile following the existing format (with frontmatter) - Test in a linked project
- Submit a PR with documentation updates
MIT License - see LICENSE file for details