One command to start everything. A complete development workflow orchestrator for AI coding agents.
English | 中文
- What is Integrated Development Workflow?
- Get Started
- Supported AI Agents
- Installation
- Usage
- Agent Compatibility
- File Structure
- Required Sub-Skills
- Zero-Dependency Hybrid Mode
- Workflow Phases
- Examples
- Troubleshooting
- License
- Contributing
Integrated Development Workflow flips the script on ad-hoc AI coding. Instead of vibe-coding every piece from scratch, this skill provides a complete, orchestrated workflow that guides the agent through:
- Requirements — Clear, documented goals
- Planning — Task breakdown with dependencies
- Implementation — TDD with proper branch management
- Testing & Review — Verification before completion
- Completion — Clean merge/PR workflow
Simply tell the agent what you want to build:
I want to build a user login feature
Implement a shopping cart
Create an API for user management
Fix the login timeout bug
Refactor the data layer
The skill will automatically:
- Check for previous session
- Create tracking files (task_plan.md, findings.md, progress.md)
- Guide through requirements
- Plan the implementation
- Execute with best practices
| Agent | Support | Notes |
|---|---|---|
| Claude Code | Full | Hooks supported |
| OpenCode | Full | Via .agents/skills |
| Cursor | Full | Via .cursor/skills |
| Trae | Full | Via .trae/skills |
| Pi Agent | Full | Via npm |
| Windsurf | Full | Via .windsurf/skills |
| Roo Code | Full | Via .roo/skills |
| Codex CLI | Full | Via .codex/skills |
| Generic | Full | Reference skill |
mkdir -p .claude/skills
cp -r /path/to/integrated-dev-workflow .claude/skills/mkdir -p .agents/skills
cp -r /path/to/integrated-dev-workflow .agents/skills/mkdir -p .cursor/skills
cp -r /path/to/integrated-dev-workflow .cursor/skills/mkdir -p .trae/skills
cp -r /path/to/integrated-dev-workflow .trae/skills/pi install npm:integrated-dev-workflowmkdir -p .windsurf/skills
cp -r /path/to/integrated-dev-workflow .windsurf/skills/mkdir -p .agents/skills
cp -r /path/to/integrated-dev-workflow .agents/skills/Simply tell the agent what you want to build. The skill automatically orchestrates the entire development process.
skill("integrated-dev-workflow")| Feature | Claude Code | OpenCode | Cursor | Trae | Pi Agent | Windsurf |
|---|---|---|---|---|---|---|
| Hooks | Full | Full | Limited | Limited | Limited | Limited |
| Session recovery | Automatic | Automatic | Via script | Via script | Via script | Via script |
| File templates | All | All | All | All | All | All |
| TDD workflow | Full | Full | Full | Full | Full | Full |
| Code review workflow | Full | Full | Full | Full | Full | Full |
This skill uses hooks for persistent reminders:
- PreToolUse: Reminds to update task_plan.md before major actions
- PostToolUse: Prompts to update task status after file changes
- Stop: Confirms task progress before ending session
Full support via .agents/skills directory:
- Copy this skill folder to
.agents/skills/ - Skills are automatically discovered
- Use
skill("integrated-dev-workflow")to invoke
Similar to Claude Code, hooks may work if Cursor supports extensions:
- Try copying to
.cursor/skills/ - If hooks not supported, manual session recovery
Trae is built on VS Code extension model:
- Copy to
.trae/skills/or VS Code extensions folder - May require manual workflow management
- Hooks are not supported in Pi Agent
- Session recovery requires manual script:
python3 scripts/session-recovery.py .
When installed, this skill creates tracking files:
your-project/
├── task_plan.md # Phase tracking, task checklist
├── findings.md # Research, decisions, notes
└── progress.md # Session log, test results, errors
# Task Plan
## Goal
[BUILD X]
## Phases
- [ ] Phase 1: Requirements
- [ ] Phase 2: Planning
- [ ] Phase 3: Implementation
- [ ] Phase 4: Testing & Review
- [ ] Phase 5: Completion
## Current Phase
Phase 1
## Tasks
- [ ] Task 1
- [ ] Task 2# Findings
## Research
- [research notes]
## Technical Decisions
- [decisions made]
## Notes
- [additional notes]# Progress
## Session Log
- Started: 2024-01-01 10:00
- Created task_plan.md
- Defined requirements with user
## Test Results
| Test | Status |
|------|--------|
| | |
## Errors Encountered
| Error | Resolution |
|-------|------------|
| | |This skill orchestrates these sub-skills:
| Skill | Purpose |
|---|---|
| planning-with-files | File-based task tracking |
| brainstorming | Requirement clarification |
| writing-plans | Task refinement |
| using-git-worktrees | Branch management |
| subagent-driven-development | Task execution |
| test-driven-development | TDD workflow |
| systematic-debugging | Issue resolution |
| verification-before-completion | Quality verification |
| requesting-code-review | Code review |
| receiving-code-review | Review handling |
| finishing-a-development-branch | Completion |
This skill works in two modes:
If sub-skills are installed, this skill delegates to them for optimal experience:
- Try
skill("brainstorming")first - If skill not found, use inline fallback guide
If sub-skills are NOT installed, this skill uses built-in fallback content:
- All core workflows are documented inline
- Templates are embedded in this skill
- You follow the same process, but with guidance in this file
- First: Try to invoke sub-skill
skill("xxx") - If not found: Use the inline fallback guide for that step
- Result: Works identically either way - just different experience levels
Recommended: Install sub-skills for best experience, but skill works completely without them!
- Define requirements with user
- Create specification (via spec-kit)
- Review and approve spec
- Plan technical approach
- Break into tasks
- Identify dependencies
- Create feature branch
- TDD for each task
- Update progress continuously
- Run all tests
- Verify build passes
- Code review
- Final verification
- Create PR / merge
- Update final status
User: "Add user authentication"
- Creates task_plan.md
- Asks: "What should auth include?"
- Documents requirements
- Plans: login, register, password reset, token handling
- Implements each with TDD
- Verifies and creates PR
User: "Fix the login timeout"
- Creates task_plan.md
- Asks: "When does it timeout?"
- Researches in findings.md
- Plans: fix timeout, add retry
- Implements and verifies
User: "Refactor data layer"
- Creates task_plan.md
- Documents current problems
- Plans: extract interface, create repo, migrate
- Executes with test coverage
- Full regression testing
Solution: Read existing files, ask user to resume or start fresh
Solution: Create minimal task_plan.md, note assumptions in findings.md
Solution: Break into phases, use subtask files
Solution: Update task_plan.md with exact next step before stopping
MIT License - See LICENSE for details.
Contributions welcome! See CONTRIBUTING.md for guidelines.
| Document | English | Chinese |
|---|---|---|
| Usage Guide | README.md | README.zh-CN.md |
| Contributing | CONTRIBUTING.md | CONTRIBUTING.zh-CN.md |
| Pressure Tests | tests/scenarios/pressure-tests.md | tests/scenarios/pressure-tests.zh-CN.md |