Summary

This repository transforms Claude Code from a single-task AI into a comprehensive engineering organization. It addresses the need for structured, disciplined AI-assisted software development by providing 12 specialized agents, 15 automation hooks, and a robust methodology. The project benefits engineers, researchers, and power users by enhancing code quality, streamlining workflows, and mitigating common development pitfalls through a team-based, rule-driven approach.

How It Works

The project structures Claude Code as a team of 12 specialized AI agents, including a planner, full-stack engineer, critic, and security verifier, each with defined roles and tool permissions governed by a strict delegation rulebook. It enforces discipline via the P7/P9/P10 methodologies and "three red lines": Closure, Fact-Driven, and Exhaustiveness. Fifteen automation hooks are integrated at critical points—pre-commit, post-tool-use, and stop events—to proactively catch issues like debugger statements, hardcoded secrets, cost overruns, and system outages before they reach production. The workflow involves task decomposition by a planner, parallel execution by engineers, and mandatory review by a critic, with security tasks potentially involving a dedicated vulnerability verifier.

Quick Start & Requirements

Installation is streamlined via the Claude Code plugin marketplace: /plugin marketplace add NYCU-Chung/my-claude-devteam followed by /plugin install devteam@my-claude-devteam. Upon restarting Claude Code, all 12 agents and 15 hooks are automatically registered. Optionally, the methodology document can be downloaded and placed in ~/.claude/CLAUDE.md for enhanced discipline. A manual installation process involves cloning the repository and copying agent and hook configurations.

Highlighted Details

• Specialized Agent Roles: Features 12 distinct agents, covering planning, full-stack development, design, security verification, debugging, database expertise, and research, each with tailored tool access.
• Rigorous Methodologies: Implements P7/P9/P10 task scoping modes and enforces "three red lines" (Closure, Fact-Driven, Exhaustiveness) for disciplined, high-quality output.
• Proactive Automation Hooks: Integrates 15 automation hooks that trigger at key development stages (e.g., pre-commit, post-tool-use) to catch common errors, security flaws, and cost issues automatically.
• Effective Code Review: The critic agent is highlighted for routinely identifying numerous bugs and security vulnerabilities in both mid-sized modules and large open-source codebases.

Maintenance & Community

The core development is attributed to @NYCU-Chung. The P7/P9/P10 methodologies and PUA mode are adapted from the MIT-licensed tanweai/pua project by Tanwei Security Lab. No specific community channels (like Discord or Slack) are listed in the README.

Licensing & Compatibility

The project is released under the MIT License, permitting broad use, modification, and distribution, including for commercial purposes and integration into closed-source projects.

Limitations & Caveats

This repository provides a methodology and agent framework, not a complete solution. Users must supply project-specific subagents for roles like VPS operations or deployment automation, configure hook paths and thresholds, and define their own CLAUDE.md project sections for security. It does not redistribute third-party skill packs, requiring users to integrate external capabilities themselves. The effectiveness is contingent on the Claude Code environment.