A collection of specialized AI agents for Claude Code designed to enhance your development workflows with intelligent automation and best practices.
This repository contains production-ready Claude Code agents that integrate seamlessly into your development environment. Each agent is designed to handle specific workflows with professional-grade quality and adherence to industry standards.
A professional Git workflow agent that analyzes code changes and generates Conventional Commits compliant commit messages with intelligent change classification.
Key Features:
- Analyzes
git diffoutput and classifies changes by package/module and functionality - Generates structured, production-ready commit messages following Conventional Commits specification
- Supports granular control over what to commit with unique identifiers for each change group
- Automatically groups related changes logically (by module, not by commit type)
- Enforces strict commit types:
feat,fix,docs,style,refactor,perf,test,chore
Usage:
# Analyze changes and generate commit messages
/commit_message
# Commit a specific change group with optional corrections
/commit <group-name> <type>(<scope>): <message>
# Commit all remaining changes automatically
/commit_allExample Workflow:
# 1. Analyze your changes
> /commit_message
# Agent analyzes diff and outputs:
# - Unique names for each logical change group
# - Classification summary organized by package/module
# - Ready-to-use commit messages
# 2. Commit specific groups
> /commit create-shared-module feat(shared): create shared module for types and utils
> /commit add-pagination feat(api): add pagination to marketplace endpoints
# 3. Commit all remaining changes at once
> /commit_allA powerful git history analysis agent that transforms raw commit data into meaningful insights by identifying temporal development clusters, tracking project evolution, and extracting patterns from your git timeline.
Key Features:
- Detects temporal commit clusters using configurable time windows (identifies activity bursts)
- Generates chronological narratives with milestone markers and development phases
- Tracks component/module evolution over time with detailed change history
- Provides statistical analysis: velocity metrics, contributor patterns, commit distributions
- Supports multiple output formats: narrative, statistics, JSON, Mermaid timelines
- Read-only operation with performance-aware defaults
Usage:
# Comprehensive analysis of recent commits
/git-commits-summarizer:analyze --since="2 weeks ago"
# Generate timeline with milestones
/git-commits-summarizer:timeline --since="1 month ago" --granularity=weekly
# Detect and analyze commit clusters
/git-commits-summarizer:clusters HEAD~100..HEAD
# Track specific component evolution
/git-commits-summarizer:evolution --path=src/authExample Workflow:
# 1. Analyze recent development activity
> /git-commits-summarizer:analyze --since="1 month ago" --limit=100
# Agent analyzes commits and provides:
# - Development phases and timeline narrative
# - Commit cluster identification (activity bursts)
# - Velocity metrics and trends
# - Breaking points detection (major changes)
# - Pattern insights and recommendations
# 2. Generate visual timeline for documentation
> /git-commits-summarizer:timeline --since="v1.0.0" --granularity=weekly
# 3. Track how a specific module evolved
> /git-commits-summarizer:evolution --path=plugins/git-commit-generator
# Agent shows:
# - Chronological evolution of the module
# - Related commits grouped by cluster
# - Change patterns and development insightsA specialized agent for recursively searching and extracting knowledge from markdown documentation by following internal links. Perfect for navigating complex documentation structures and finding specific information across interconnected files.
Key Features:
- Intelligent entry point detection (
.llms-txt,docs/,README.md) - Recursive link traversal with circular reference prevention
- Relevance-based result ranking (High/Medium/Low)
- Context extraction with link chain tracking
- Read-only operation ensuring safe documentation exploration
- Fallback strategies with user confirmation for expanded searches
Usage:
# Search with specific entry file
/research authentication API in docs/api/readme.md
# Search from default documentation root
/research database migrations
# Search with quoted phrases
/research "error handling" in docs/guide.mdExample Workflow:
# 1. Search for specific topic
> /research JWT token validation in docs/security/auth.md
# Agent recursively follows markdown links and returns:
# - Relevance-ranked findings
# - Source paths with link chains
# - Search statistics (files searched, links followed)
# 2. If nothing found, agent asks for permission to expand search
# Agent: "No results found. Search from root documentation?"
> Yes
# 3. Agent expands search to .llms-txt and presents comprehensive resultsA meta-agent that guides you through creating new agents for this repository. Streamlines the agent development workflow with automated scaffolding and quality checks.
Key Features:
- Interactive requirements gathering
- Automated file generation with proper YAML frontmatter
- Marketplace.json registration
- Consistency validation across all references
- Follows established patterns from existing agents
Usage:
# Launch the agent creator
Can you help me create a new agent for [specific purpose]?
# Or use the slash command
/create_agentA specialized agent for creating custom Claude Code skills with proper structure, YAML frontmatter, and best practices guidance. Skills are autonomously-invoked capabilities that extend Claude's functionality.
Key Features:
- Interactive skill creation workflow
- Clear guidance on skills vs agents (when to use each)
- Proper YAML frontmatter and SKILL.md structure
- Support for reference files (REFERENCE.md) and executable code
- Both personal (~/.claude/skills/) and project (.claude/skills/) skill support
- Comprehensive validation and consistency checks
Usage:
# Launch the skill creator
/create_skill
# Or natural request
Can you help me create a skill for [specific purpose]?Example Workflow:
# 1. Create a new skill
> /create_skill
# Agent asks about your requirements:
# - What's the skill's purpose?
# - What domain knowledge should it provide?
# - Should it be personal or project-specific?
# 2. Agent generates skill structure
# .claude/skills/{skill-name}/
# ├── SKILL.md # Main skill file
# ├── REFERENCE.md # (Optional) detailed reference
# └── [code-files] # (Optional) executable scripts
# 3. Skill is ready to use - Claude will invoke it automatically
# when relevant based on the skill's description# Clone the repository
git clone https://github.com/aztack/claude-code-agents.git
# Link to Claude Code agents directory (adjust path as needed)
ln -s $(pwd)/aztack-claude-code-agents ~/.claude/agents/aztack-claude-code-agents(Coming soon - once published to the official marketplace)
Search for "aztack-claude-code-agents" or "git-commit-generator" in the Claude Code marketplace and click install.
aztack-claude-code-agents/
├── .claude/
│ ├── agents/
│ │ └── agent-creator.md # Meta-agent for creating new agents
│ └── commands/
│ └── create_agent.md # Slash command for agent creation
├── .claude-plugin/
│ └── marketplace.json # Plugin marketplace configuration
├── plugins/
│ ├── git-commit-generator/
│ │ ├── agents/
│ │ │ └── git-commit-generator.md
│ │ └── commands/
│ │ ├── commit.md
│ │ ├── commit_all.md
│ │ └── commit_message.md
│ ├── git-commits-summarizer/
│ │ ├── agents/
│ │ │ └── git-commits-summarizer.md
│ │ └── commands/
│ │ ├── analyze.md
│ │ ├── timeline.md
│ │ ├── clusters.md
│ │ └── evolution.md
│ ├── knowledge-researcher/
│ │ ├── agents/
│ │ │ └── knowledge-researcher.md
│ │ └── commands/
│ │ └── research.md
│ ├── skill-creator/
│ │ ├── .claude-plugin/
│ │ │ └── plugin.json
│ │ ├── agents/
│ │ │ └── skill-creator.md
│ │ └── commands/
│ │ └── create_skill.md
│ └── d2c-generator/
│ └── ...
├── CLAUDE.md # Development guidance for Claude Code
└── README.md
Automated Approach (Recommended):
Use the agent-creator meta-agent to streamline the process:
# In Claude Code conversation
Can you help me create a new agent for [specific purpose]?The agent-creator will:
- Gather requirements interactively
- Generate proper file structure with YAML frontmatter
- Register the agent in marketplace.json
- Validate consistency across all references
Manual Approach:
-
Create a new directory under
plugins/with kebab-case naming:mkdir -p plugins/my-new-agent/{agents,commands} -
Create the agent definition file with YAML frontmatter:
--- name: my-new-agent description: | Agent description with usage examples... model: inherit --- [Agent system prompt and instructions here]
-
Create command files if needed (optional):
# plugins/my-new-agent/commands/my_command.md --- description: Command description argument-hint: "<args>" --- [Command implementation]
-
Update
.claude-plugin/marketplace.jsonto register the plugin:{ "plugins": [ { "name": "my-new-agent", "source": "./plugins/my-new-agent", "agents": ["./agents/my-new-agent.md"], "commands": ["./commands/my_command.md"] } ] }
Test agents directly in Claude Code using the Task tool:
# In Claude Code conversation
Can you use the git-commit-generator agent to analyze my changes?
# Or test slash commands
/research "authentication" in docs/Contributions are welcome! Please follow these guidelines:
- Follow the existing agent structure and naming conventions
- Ensure YAML frontmatter is valid
- Include comprehensive examples in the agent description
- Test agents thoroughly before submitting
- Update marketplace.json with appropriate metadata
MIT License - see LICENSE file for details
Wang Weihua (aztack)
- GitHub: @aztack
- Email: aztack@163.com
- Added skill-creator plugin for creating custom Claude Code skills
- Interactive workflow for skill creation with proper structure
- Clear guidance on skills vs agents (when to use each)
- Support for SKILL.md, REFERENCE.md, and executable code files
- Both personal (~/.claude/skills/) and project (.claude/skills/) skill support
- Comprehensive validation and best practices enforcement
/create_skillslash command for quick access
- Enhanced documentation with skill creation examples and workflows
- Added git-commits-summarizer plugin for comprehensive git history analysis
- Temporal commit cluster detection with configurable time windows
- Chronological timeline narratives with milestone markers
- Component evolution tracking with detailed change history
- Statistical analysis: velocity metrics, contributor patterns, distributions
- Multiple output formats: narrative, statistics, JSON, Mermaid diagrams
- Four specialized commands: analyze, timeline, clusters, evolution
- Added /create_agent slash command for streamlined agent creation workflow
- Enhanced repository structure with .claude/commands/ directory
- Added knowledge-researcher agent for recursive documentation search
- Intelligent entry point detection and link traversal
- Relevance-based ranking with context extraction
- Read-only safe operation with fallback strategies
- Added agent-creator meta-agent for streamlined agent development
- Interactive requirements gathering
- Automated scaffolding and validation
- Marketplace registration and consistency checks
- Reorganized repository structure with
plugins/directory - Enhanced development workflow documentation
- Added git-commit-generator agent with full Conventional Commits support
/commit_message- Analyze and generate commit messages/commit- Commit specific change groups/commit_all- Bulk commit all remaining changes
- Established repository structure and development guidelines
- Created marketplace configuration for distribution
Note: These agents are designed to work with Claude Code CLI. For more information about Claude Code, visit claude.com/claude-code.