详细介绍
Detailed Introduction
Repository purpose: a ready-to-deploy “AI sub-agent system” template library, organized by functional departments and covering full software workflows such as design, engineering, product, marketing, operations, testing, and project management. The contents are standalone agent definition files (Markdown with YAML frontmatter) that can be called by Claude Code and copied directly into ~/.claude/agents/ for immediate use.
Brief description (concise, for technical readers):
This repository gathers a set of plug-and-play professional AI agents organized to mirror the roles in real software teams, covering the main stages of the product development lifecycle. Each agent is defined as a Markdown file including YAML configuration, role definition, usage examples, tool permissions, and system prompts, which Claude Code can automatically recognize and invoke.
Core Values
- Standardization: unified agent templates with repeatable and extensible structure.
- Specialization: organized by departments such as engineering, product, design, and marketing.
- Automation: Claude Code automatically triggers the most relevant agent based on context.
- Portability: copy directly to the local
~/.claude/agents/directory to load. - Extensibility: includes a complete guide for adding new agents (templates, field descriptions, example requirements).
Directory Overview
- engineering: AI engineer, backend architecture, DevOps automation, frontend development, rapid prototyping, etc.
- product: trend research, feedback synthesis, prioritization, etc.
- design: UI/UX design, brand consistency, visual storytelling, etc.
- marketing: short-video trends, growth strategy, content creation, community operations, etc.
- project-management: experiment tracking, delivery management, producer roles, etc.
- studio-operations: analytics, finance, legal, infrastructure, customer support, etc.
- testing: API testing, performance benchmarking, result analysis, tool evaluation, workflow optimization.
- bonus: coaching and recreational agents.
Installation
git clone https://github.com/contains-studio/agents.git
cp -r agents/* ~/.claude/agents/
Claude Code will automatically load the agents after a restart.
Typical Usage
No explicit invocation is required — describe the task scenario and Claude Code will pick the appropriate agent. Examples:
- “Build an MVP for a meditation habit-tracking app” → rapid-prototyper
- “User negative feedback spiked, summarize the core issues” → feedback-synthesizer
- “Optimize App Store keyword rankings” → app-store-optimizer
- “Write an end-to-end test suite for a new feature” → test-writer-fixer
Extension Guide
The repository provides full agent templates, including:
- YAML frontmatter (name, description, examples, color, tool permissions)
- System prompt structure of ≥500 words
- Example requirements for different departments (e.g., engineering agents require: feature implementation, bug fixes, refactoring, architecture decisions)
- Trigger rules, quality validation, and performance/effectiveness evaluation methods
Intended Audience
- Developers building AI-native workflows
- Teams that need multi-agent collaboration
- Studios that want to quickly build internal AI assistant systems
- Individual engineers learning professional sub-agent system design
Summary
This is a practical, deployable AI agent team template library that enables individuals or teams to quickly adopt a multi-role AI collaboration workflow. It can be extended, customized, and embedded into real development processes as needed.