Agent Workflow Template: A Standardized Workflow Template for AI-Assisted Development

Agent Workflow Template is a GitHub template repository designed to help developers quickly establish standardized workflows for AI-assisted development. It addresses the problem of teams lacking unified collaboration guidelines when using AI coding assistants. Its core components include AGENTS.md collaboration guidelines, Cursor rules and skills, Claude Code commands, Husky hooks, and a complete CI/CD process, providing a reusable and extensible collaboration framework for AI-assisted development.

With the popularity of AI coding assistants like GitHub Copilot, Cursor, and Claude Code, the way developers collaborate with AI has changed. However, teams face the challenge of lacking standardized collaboration guidelines: each developer has different usage habits, leading to inconsistent code quality and AI output. As a meta-level framework, this template helps teams maintain consistent development experiences and quality standards by defining clear rules, configurations, and workflows.

AGENTS.md - Core of Collaboration Guidelines

Contains Learned Workspace Facts (project tech stack, architecture decisions, etc.) and Learned User Preferences (developer habits, code style, etc.), serving as both a project onboarding manual and an AI memory bank.

Cursor Rules and Skills

• Rule files: The .cursor/rules/ directory contains 12 types of guidelines such as always.md and agentic-workflow.mdc
• Skill library: The .cursor/skills/ directory includes 4 skills like code-review and deslop

Claude Code Commands

• /loop-on-ci: Continuously monitors PR CI status and automatically fixes failed items
• /vercel-setup: One-click configuration of Vercel deployment environment
• /agent-os-bootstrap and /agent-os-retro: Support a seven-layer personal agent operating system

GitHub Actions Workflows

Preconfigured with 6 types of workflows including agentic-ci.yml (code inspection) and ai-review.yml (AI review)

Husky Hooks

• pre-commit: Prettier formatting → ESLint → TypeScript check → Vitest testing
• pre-push: Document integrity check

CodeRabbit Configuration

Uses a "chill" tone, sets the docstring coverage warning threshold at 80%, and includes custom deslop and nl-check checks

Initialize a New Project

1. Create a new repository using the template
2. Update Workspace Facts in AGENTS.md
3. Ensure package.json includes scripts like lint and format
4. Activate Git hooks (husky install / npm run prepare)
5. Configure CodeRabbit
6. Set up keys like OPENAI_API_KEY
7. Run /vercel-setup (if Vercel deployment is needed)
8. Run /agent-os-bootstrap to initialize the agent system

Daily Development Workflow

1. Coding: Cursor/Claude automatically apply rules
2. Before commit: Husky automatically runs formatting, linting, and testing
3. PR creation: CodeRabbit/Copilot automatically review
4. CI phase: agentic-ci.yml verifies quality
5. Continuous fixing: /loop-on-ci monitors and fixes CI failures
6. After merge: post-merge-verify.yml verifies the main branch

1. Multi-agent Collaboration: Cursor (daily coding), Claude (complex tasks), CodeRabbit (review), Copilot (completion), and Husky (local checks) each perform their respective roles
2. Rules as Code: Guidelines are stored in Markdown/YAML, enabling version control, team sharing, AI understandability, and testability
3. Progressive Adoption: No specific application code/framework configurations; developers enable components as needed, lowering the entry barrier

Applicable Scenarios

• Newly launched projects: Establish AI collaboration guidelines
• Team collaboration projects: Unify AI tool usage
• Open-source projects: Provide contributor collaboration guidelines
• AI-native development: Teams centered on AI collaboration

Target Users

• Technical teams standardizing AI development workflows
• Project maintainers pursuing code quality
• Developers looking to reduce review friction
• Early adopters of AI-Native development models

1. Tech stack assumptions: Some hooks/workflows are based on Node.js + Next.js; non-Node projects need adjustments
2. Experimental features: .cursor/hooks.ts is experimental and needs updates with Cursor versions
3. Command compatibility: /loop-on-ci is designed specifically for Claude Code CLI
4. Learning curve: First-time users need to understand component collaboration; it's recommended to expand gradually from core features

Agent Workflow Template promotes the evolution of AI-assisted development toward standardization and engineering, establishing a collaborative partnership between humans and AI based on clear contracts. It helps teams maintain controllable code quality while enjoying AI efficiency, serving as an important starting point for exploring AI-Native development models. As AI coding assistants evolve, the value of such templates will become increasingly prominent.