Skip to content
/
Tier B A6 Multi-agent orchestrator

Claude Code Templates (davila7)

claude-code-templates-davila · davila7/claude-code-templates · ★ 28k · last commit 2026-05-26

Primitive shape 24 total
Commands 8 Subagents 15 Hooks 1
00

Summary

Claude Code Templates (davila7) — Summary

claude-code-templates by Daniel Avila is the largest component-catalog framework in this batch: 27,587 stars, 30 contributors, actively maintained. It ships a Node.js CLI (npx claude-code-templates@latest) for installing individual or batch components into any project, plus a web UI at aitmpl.com for browsing ~1,000+ components. The repo contains 15 agents, 8 commands, 1 hook script, and a dashboard web application (Astro/Tailwind). Components span agents (frontend-developer, security-auditor, etc.), commands (worktree management with Ghostty integration, lint, test), MCPs (GitHub, PostgreSQL, Stripe, AWS, OpenAI integrations), settings, hooks, and skills. The project is sponsored by Z.AI and Anthropic's Open Source Program. The analytics feature (--analytics flag) provides real-time Claude Code session monitoring with live state detection and performance metrics. The Telegram webhook hook (telegram-pr-webhook.py) demonstrates external notification integration.

Differs from seeds: Closest to openspec (Archetype 2 — component marketplace with install commands) but operates as a component store rather than a workflow framework. Unlike bmad-method (34 skills, behavioral enforcement via personas), davila7 ships isolated, mix-and-match components the user assembles. Unlike taskmaster-ai (MCP + task management), davila7 is a distribution platform for Claude Code primitives of all types. The worktree commands with Ghostty-panel integration are unique in this batch.

01

Overview

Claude Code Templates (davila7) — Overview

Origin

Created by Daniel Avila (@daniavila). 27,587 stars, 30 contributors, MIT license. Sponsored by Z.AI and Anthropic's Claude for Open Source Program. Listed on Trendshift. Last commit 2026-05-26 (actively maintained).

Mission

"Ready-to-use configurations for Anthropic's Claude Code. A comprehensive collection of AI agents, custom commands, settings, hooks, external integrations (MCPs), and project templates to enhance your development workflow."

Philosophy

Component marketplace model: users install only what they need. The npm CLI lets developers compose their own Claude Code setup from battle-tested community components. No opinion about workflow structure — pure toolbox.

Component count

From CLAUDE.md (internal count, not just the .claude/ directory):

  • Agents: 600+
  • Commands: 200+
  • MCPs: 55+
  • Settings: 60+
  • Hooks: 39+
  • Templates: 14+

Note: The .claude/ directory in the repo root contains only the project's own working configuration (15 agents, 8 commands, 1 hook). The bulk of the catalog is browsed via the web dashboard and installed via the CLI.

Dashboard

Interactive web UI at https://www.aitmpl.com (beta) — browse and install components. Also deployable as a Claude Code analytics dashboard.

Unique features

  1. Analytics mode: npx claude-code-templates@latest --analytics — live Claude Code session monitoring
  2. Worktree + Ghostty integration — commands that create parallel worktrees and set up Ghostty terminal panels for simultaneous Claude instances
  3. Telegram webhook hook — PR notifications via Telegram bot
  4. Neon DB integration (Open Source Program partnership)
02

Architecture

Claude Code Templates (davila7) — Architecture

Distribution

  • CLI: npx claude-code-templates@latest (Node.js npm package)
  • Dashboard: https://www.aitmpl.com (Astro + Tailwind, Vercel-hosted, Cloudflare Workers)
  • Supabase for analytics/download tracking
  • Vercel API endpoints for download tracking + Discord integration

Install

# Interactive browser
npx claude-code-templates@latest

# Specific components
npx claude-code-templates@latest --agent frontend-developer --yes
npx claude-code-templates@latest --command testing/generate-tests --yes
npx claude-code-templates@latest --hook git/pre-commit-validation --yes
npx claude-code-templates@latest --mcp database/postgresql-integration --yes

# Batch install
npx claude-code-templates@latest --agent development-team/frontend-developer --command testing/generate-tests --mcp development/github-integration --yes

Project directory (working config)

.
├── .claude/
│   ├── agents/          (15 agents)
│   ├── commands/        (8 commands)
│   ├── hooks/
│   │   └── telegram-pr-webhook.py
│   ├── rules/
│   └── launch.json
├── .mcp.json            (MCP server configs)
├── dashboard/           (Astro web app)
│   └── src/
├── api/                 (Vercel functions)
├── cli-tool/            (CLI source)
├── cloudflare-workers/
├── database/
├── scripts/
│   └── generate_components_json.py
└── docs/
    └── components.json  (catalog index)

Runtime requirements

  • Node.js (CLI + dashboard)
  • Python (for generate_components_json.py catalog generator)
  • Vercel (for API + hosting)

Target AI tools

  • Claude Code (primary, .claude/ configuration)
  • Any MCP client (for MCP components)
03

Components

Claude Code Templates (davila7) — Components

Agents (15 in repo's .claude/agents/)

Name Purpose
agent-expert.md Claude Code agent development specialist
blog-writer.md Blog content creation
build-checker.md Build validation specialist
catalog-generator.md Component catalog generation
cli-ui-designer.md CLI/UI design specialist
command-expert.md Claude Code command development
component-improver.md Component quality improvement
component-migrator.md Component migration between formats
component-researcher.md Component research and analysis
component-reviewer.md Component review and validation
deployer.md Deployment automation specialist
docusaurus-expert.md Docusaurus documentation specialist
frontend-developer.md React/responsive design specialist (model: sonnet)
linear-tracker.md Linear issue tracking integration
mcp-expert.md MCP server development specialist

Commands (8 in .claude/commands/)

Name Purpose
cleanup-cache.md Cache cleanup
create-blog-article.md Blog article creation workflow
lint.md Run linter
test.md Run tests
worktree-check.md Check worktree status
worktree-cleanup.md Remove worktrees
worktree-deliver.md Deliver work from worktree (merge/PR)
worktree-init.md Create parallel worktrees for Ghostty panels

Hooks (1 script)

Name Language Purpose
telegram-pr-webhook.py Python Send PR notifications to Telegram

MCP (.mcp.json)

External MCP server configurations (not bundled servers; references to external MCPs):

  • GitHub integration
  • Possibly others (not fully fetched)

CLI features

Flag Purpose
--agent <name> Install specific agent
--command <name> Install specific command
--hook <name> Install specific hook
--mcp <name> Install MCP server config
--setting <name> Install setting
--yes Skip confirmation prompts
--analytics Launch analytics dashboard

Catalog (community, not in repo)

  • 600+ agents
  • 200+ commands
  • 55+ MCPs
  • 60+ settings
  • 39+ hooks
  • 14+ templates
05

Prompts

Claude Code Templates (davila7) — Prompts

Prompt 1: frontend-developer agent

Source: .claude/agents/frontend-developer.md

Technique: Role persona with focus areas + output contract

---
name: frontend-developer
description: Frontend development specialist for React applications and responsive design. Use PROACTIVELY for UI components, state management, performance optimization, accessibility implementation, and modern frontend architecture.
tools: Read, Write, Edit, Bash
model: sonnet
---

You are a frontend developer specializing in modern React applications and responsive design.

## Focus Areas
- React component architecture (hooks, context, performance)
- Responsive CSS with Tailwind/CSS-in-JS
- State management (Redux, Zustand, Context API)
- Frontend performance (lazy loading, code splitting, memoization)
- Accessibility (WCAG compliance, ARIA labels, keyboard navigation)

## Approach
1. Component-first thinking — reusable, composable UI pieces
2. Mobile-first responsive design
3. Performance budgets — aim for sub-3s load times
4. Semantic HTML and proper ARIA attributes
5. Type safety with TypeScript when applicable

## Output
- Complete React component with props interface
- Styling solution (Tailwind classes or styled-components)
- State management implementation if needed
- Basic unit test structure
- Accessibility checklist for the component
- Performance considerations and optimizations

Focus on working code over explanations. Include usage examples in comments.

Key design: "Use PROACTIVELY" signals to Claude Code this is an auto-delegation agent, not a reactive one. Explicit output contract defines the deliverables.

Prompt 2: worktree-init command (excerpt)

Source: .claude/commands/worktree-init.md

Technique: Sequential procedural workflow with validation gates

---
allowed-tools: Bash(git:*), Bash(mkdir:*), Bash(ls:*), Bash(cat:*), Bash(basename:*), Bash(pwd:*), Bash(sed:*)
argument-hint: task 1 | task 2 | task 3
description: Create parallel worktrees for multi-task development with Ghostty panels
---

Create multiple git worktrees for parallel development: $ARGUMENTS

### Step 1: Validate Environment
1. Check this is a git repository: `git rev-parse --is-inside-work-tree`
2. Get the repo name: `basename $(git rev-parse --show-toplevel)`
3. Ensure working tree is clean: `git status --porcelain`. If dirty, warn the user and ask if they want to continue.
4. Fetch latest: `git fetch origin`

Key design: allowed-tools restricts Bash to specific subcommand patterns — precise capability scoping to prevent unintended side effects. Ghostty-specific terminal setup is unique in this corpus.

09

Uniqueness

Claude Code Templates (davila7) — Uniqueness

Differs from seeds

The closest seed archetype is openspec (Archetype 2 — component store with install commands), but davila7 is more ambitious: it's a full-stack component marketplace with a web dashboard, npm CLI, analytics, and community-maintained catalog of 600+ agents. Unlike bmad-method or superpowers, which enforce a specific behavioral framework, davila7 is opinionated about nothing — it's a distribution platform. The worktree-init + Ghostty integration is the only implementation in the corpus that creates parallel Claude Code sessions linked to terminal panel management. The Telegram webhook hook is the only outbound messaging integration in this batch.

Positioning

The npm "component store" for Claude Code — the AppStore model applied to AI agent primitives. High volume, broad coverage, no forced workflow structure. Community-maintained; 30 contributors.

Observable failure modes

  1. No workflow coherence: installing 10 components from different authors produces no coherent workflow — the user must assemble their own orchestration.
  2. Versioning across 600+ components: maintaining quality at scale with 30 contributors is challenging; component compatibility across Claude Code versions is not guaranteed.
  3. Dashboard beta: aitmpl.com is in beta; reliability for production use is unknown.
  4. Ghostty dependency: worktree-init.md assumes Ghostty as the terminal emulator — users on iTerm2, Terminal.app, or Linux terminals need to adapt.
  5. Hardcoded local paths: The .env.example and secret management warnings suggest past incidents with committed credentials.
04

Workflow

Claude Code Templates (davila7) — Workflow

User workflow: install + use

  1. Browse components at aitmpl.com or via npx claude-code-templates@latest (interactive)
  2. Install desired components: npx claude-code-templates@latest --agent <name> --command <name>
  3. Components are placed into .claude/agents/, .claude/commands/, .claude/hooks/, etc.
  4. Use installed components within Claude Code sessions

Developer workflow: worktree parallel development

The worktree-init.md command enables a specific multi-task workflow:

  1. Describe tasks separated by |: task 1 | task 2 | task 3
  2. Command creates git worktrees for each task
  3. Sets up Ghostty terminal panels (one per worktree)
  4. Each panel runs an independent Claude Code session
Phase Artifact
Parse tasks from args Task names
Validate git environment Clean working tree
Create worktrees ../repo-name-task-1/, etc.
Set up Ghostty panels Terminal panels per worktree
Parallel execution Independent Claude sessions

Approval gates

  • worktree-init.md: if dirty working tree → prompt user before continuing
  • Default --yes flag suppresses confirmation in batch mode

Build/release workflow (project internal)

python scripts/generate_components_json.py  # update catalog
npm version patch|minor|major
npm publish
vercel --prod
06

Memory Context

Claude Code Templates (davila7) — Memory & Context

State storage

  • CLAUDE.md (project-level) — defines component types, installation patterns, security guidelines (no hardcoded secrets)
  • docs/components.json — catalog index generated by generate_components_json.py; this is the framework's primary state file
  • Supabase — download tracking and analytics (cloud)
  • Vercel — API state for Discord integration

No persistent agent memory

Individual components are stateless; no SQLite, vector DB, or session-based memory.

Security memory (CLAUDE.md enforced)

"CRITICAL: NEVER Hardcode Secrets or IDs — NEVER write API keys, tokens, passwords, project IDs in code. ALL must go in .env."

The CLAUDE.md security guidelines are the primary memory artifact for the project's development team.

Context scope

Components are self-contained markdown files; they bring their own context when installed. No global context injection mechanism.

07

Orchestration

Claude Code Templates (davila7) — Orchestration

Multi-agent

Yes — the worktree-init workflow creates separate Claude Code instances per worktree (one per Ghostty panel), enabling parallel independent agents.

Orchestration pattern

parallel-fan-out (for worktree multi-task workflow); none (for individual component use).

Isolation mechanism

git-worktree — the worktree-init.md command creates git worktrees; each runs an independent Claude Code session. worktree-deliver.md handles merging work back.

Subagent definition format

persona-md — each agent in .claude/agents/ is a markdown file with YAML frontmatter.

Multi-model

Yes — the frontend-developer agent explicitly sets model: sonnet. Other agents may set different models (not all fetched).

Execution mode

one-shot per component invocation; the worktree workflow enables parallel sessions.

Auto-validators

From agents and commands:

  • build-checker.md agent — dedicated build validation
  • lint.md command — linting
  • test.md command — test runner

Consensus

None.

08

Ui Cli Surface

Claude Code Templates (davila7) — UI & CLI Surface

Dedicated CLI binary

Yes — npx claude-code-templates@latest (Node.js npm package, thin installer wrapper)

  • Package: claude-code-templates
  • Is thin wrapper: Yes — installs components into user's .claude/ dir; does not run Claude itself
  • Key flags: --agent, --command, --hook, --mcp, --setting, --yes, --analytics
  • Interactive mode: Prompts for component selection when no flags provided

Local web dashboard

Yes — hosted at https://www.aitmpl.com (Vercel, Astro + Tailwind)

  • Features: Browse components, manage collections, track installations
  • Status: Beta
  • Not self-hosted (no local port)

Analytics dashboard

npx claude-code-templates@latest --analytics

  • Real-time Claude Code session monitoring
  • Live state detection
  • Performance metrics

IDE integration

Claude Code (.claude/ directory)

Observability

  • Download tracking via Vercel API + Supabase
  • Discord integration for community notifications
  • telegram-pr-webhook.py — Telegram bot PR notifications

MCP reference configs (.mcp.json)

External MCP server connection configs for:

  • GitHub
  • (others not confirmed)

Related frameworks

same archetype · same primary tool · same memory type

Claude-Flow / Ruflo ★ 55k
A6 Multi-agent orchestrator

Eliminates single-agent context limits and sequential bottlenecks by orchestrating fault-tolerant swarms of specialized AI agents…

Hermes Agent (NousResearch) ★ 168k
A6 Multi-agent orchestrator

Self-improving personal AI agent with closed learning loop, 7 terminal backends, and messaging gateway — not tied to any AI…

OpenCode ★ 165k
A6 Multi-agent orchestrator

Terminal-first AI coding agent with multi-model routing, native desktop app, and a typed .opencode/ configuration system for…

OpenHands ★ 75k
A6 Multi-agent orchestrator

Open-source AI software development platform (open-source Devin alternative) with Docker sandbox isolation, 77.6% SWE-bench…

DeerFlow ★ 70k
A6 Multi-agent orchestrator

Long-horizon superagent that researches, codes, and creates by orchestrating parallel sub-agents with isolated contexts in Docker…

oh-my-openagent (omo) ★ 60k
A6 Multi-agent orchestrator

Multi-provider AI agent orchestration for OpenCode: escape vendor lock-in by routing Sisyphus (Claude/Kimi/GLM) and Hephaestus…