AI_PROJECT_TEMPLATE：面向多智能体开发的 AI 项目脚手架

AI_PROJECT_TEMPLATE is a complete AI project scaffold developed by sprustonlab, built on copier template engine. It solves common pain points in AI-assisted development (e.g., complex environment configuration, messy permission management, lack of multi-agent collaboration norms) and provides a one-click initialization for production-grade AI dev environments. Key highlights include deep support for multi-agent workflows via integration with claudechic (a TUI wrapper for Claude Code), layered permission guardrails, and a full toolchain covering project initialization to continuous development.

The template addresses critical issues faced when using AI tools like Claude Code: environment setup complexity, unregulated permission management, and absence of standardized multi-agent collaboration processes. It aims to provide a structured framework to streamline AI-assisted development from individual to team scenarios.

The template features a clear directory structure (e.g., .claude/ for Claude Code configs, workflows/ for collaboration flows, global/ for project-wide rules). Key components include:

1. claudechic: A TUI tool offering shared permission modes (default/edit/plan/bypass), session management (/clearui command), and Yolo mode (unrestricted command execution).
2. Layered Permission System: Guardrail rules (block dangerous operations), global rules (global/rules.yaml), and workflow-specific rules ensure security and flexibility.

The project-team workflow (triggered via /project-team) follows 4 stages:

1. Vision: Clarify user requirements to align all agents on project goals.
2. Specification: Lead agents draft detailed specs covering composability, terminology, user alignment, and risk identification (via Skeptic agent).
3. Implementation: Executor agents write code under lead agent guidance.
4. Testing: Run tests and get lead agent approval, with user checkpoints at each stage.

Installation options:

• One-click: Single-line commands for Linux, macOS, Windows (from project website).
• Pixi-based: Use pixi exec to copy the template and install dependencies. Configuration options during setup: project name, quick start level, target platform, claudechic mode (standard/developer), cluster support (LSF/SLURM), git initialization, and existing codebase integration.

The template supports:

• MCP Tools: Add Python plugins to mcp_tools/ for automatic loading.
• Cluster Support: Built-in LSF/SLURM integration for HPC environments.
• Developer Mode: Editable claudechic setup and access to .claude/rules/*.md (Claude Code context files).
• Version Updates: Use copier update to keep the template up-to-date while preserving custom changes.

The template is suitable for:

• Personal Projects: Out-of-box AI dev environment.
• Team Collaboration: Standardized structure reduces friction.
• Education: Teaching AI-assisted development best practices.
• Research: HPC support for AI training jobs.
• Enterprise: Compliance with security and permission requirements.

AI_PROJECT_TEMPLATE represents the maturation of AI-assisted development toolchains. It goes beyond basic code generation to provide a full engineering framework covering environment management, permission control, and collaborative workflows. It serves as a solid starting point for individuals and teams looking to integrate AI tools into their formal development processes.