Skip to content

stacmv/planning-framework

Repository files navigation

Planning Framework v2.0

Issue-based workflow for AI-assisted development across sessions and branches

License: MIT Version


What's New in v2.0

✨ Issue-Based Workflow - Each task gets its own folder with complete context 🎯 Bounded File Sizes - Global files stay small, execution details in issues 🚫 No Merge Conflicts - Issue folders are branch-specific 🤖 Multi-Agent Support - Single config for Claude/Gemini/Qwen ✅ Quality Gates - Built-in QA workflow

Upgrading from v1.0?

See MIGRATION-GUIDE.md for step-by-step instructions.


The Problem

When working with AI agents across multiple sessions:

  • ❌ Context is lost between sessions
  • ❌ Planning files grow unbounded
  • ❌ Feature branches conflict on shared planning docs
  • ❌ No structured way to track features/bugs

The Solution

Planning Framework v2.0 uses an issue-based workflow:

  • ✅ Each task in its own issue folder
  • ✅ Complete context (prompt, analysis, plan, progress)
  • ✅ Global files stay minimal (roadmap only)
  • ✅ Zero merge conflicts (branch-specific issues)
  • ✅ Multi-agent support (Claude Code, Gemini CLI, Qwen Code)

Quick Start (5 Minutes)

For New Projects

# 1. Clone planning framework
git clone https://github.com/[your-org]/planning-framework
cd planning-framework

# 2. Run interactive setup
./scripts/setup-planning-v2.sh
# Follow prompts: project name, issue types, QA requirements

# 3. Commit the framework
git add .
git commit -m "Setup Planning Framework v2.0"

# 4. Create first issue
# Ask your AI agent: "Create an issue to [add feature]"

For Existing v1.0 Projects

# Migrate from v1.0 to v2.0
./scripts/migrate-v1-to-v2.sh
# Backs up v1.0, creates v2.0 structure, generates report

See QUICKSTART.md for complete 5-minute guide.


How It Works

Issue-Based Workflow

USER REQUEST → CREATE ISSUE → ANALYZE → PLAN → IMPLEMENT → QA → CLOSE

File Structure

your-project/
├── PLANNING.md                      # Framework config
├── .qa-workflow.md                  # QA requirements
│
├── docs/
│   ├── issues/
│   │   ├── open/                    # Active work
│   │   │   └── 20240127-feat-add-auth/
│   │   │       ├── prompt.md        # Original request
│   │   │       ├── analysis.md      # Understanding
│   │   │       ├── implementation-plan.md  # Tasks
│   │   │       └── session-log.md   # Progress
│   │   │
│   │   └── closed/                  # Completed (archived)
│   │
│   └── planning/
│       ├── implementation-plan.md   # Roadmap (stays small!)
│       ├── session-log.md           # Timeline
│       └── decisions.md             # Decisions

What Agents Do

Session Start:

  1. Read PLANNING.md - Framework instructions
  2. Read issue files - Context for current task
  3. Read global decisions.md - Architectural choices

During Work:

  • Focus on ONE issue per session
  • Update progress in session-log.md
  • Check off tasks in implementation-plan.md

Before Closing:

  • Run QA workflow (.qa-workflow.md)
  • Get user confirmation
  • Merge and archive issue

Key Features

1. Issue Naming

Format: YYYYMMDD-{type}-{slug}

Example: 20240127-feat-add-authentication

  • Date ensures uniqueness across branches
  • Type visible at glance (feat/bug/improve)
  • Slug provides readability

2. Branches

One branch per issue: issue/YYYYMMDD-{type}-{slug}

  • 1:1 mapping with issue folder
  • Clean isolation, no conflicts

3. Global Files (Stay Small!)

implementation-plan.md - Roadmap + active issues session-log.md - One-line entries when issues close decisions.md - Architectural decisions

Details live in issue folders, not global files.

4. Quality Assurance

.qa-workflow.md defines gates before closing:

  • Linting/formatting
  • Tests (existing + new)
  • Documentation
  • Security review

Documentation


Multi-Agent Support

Works with:

  • Claude Code - Anthropic's CLI tool
  • Gemini CLI - Google's CLI tool
  • Qwen Code - Qwen's CLI tool
  • Any AI agent - Generic instructions

Single PLANNING.md for all agents.

Session logs track which agent did what:

[Claude Code] ✓ [issue-id](link) - Description
[Gemini CLI] 2024-01-28: Ad-hoc work

Scripts & Automation

Setup:

./scripts/setup-planning-v2.sh      # Interactive setup

Migration:

./scripts/migrate-v1-to-v2.sh       # Migrate from v1.0

Helpers:

./scripts/create-issue.sh           # Manual issue creation
./scripts/close-issue.sh            # Issue closure automation

All scripts:

  • Interactive prompts
  • Colorized output
  • Error handling
  • Comprehensive help

Benefits

For Developers

Zero context loss - Pick up exactly where you left off ✅ Clear progress - Visual tracking with checkboxes ✅ Isolated work - Issues don't interfere with each other ✅ No conflicts - Branch-specific planning files

For AI Agents

Focused context - Read only relevant issue ✅ Clear workflow - 6-phase lifecycle ✅ Consistent behavior - Follows documented patterns ✅ Reduced tokens - Small files, not 1000s of lines

For Teams

Parallel development - No merge conflicts ✅ Issue tracking - Built-in feature/bug management ✅ Quality gates - QA workflow before merging ✅ Decision history - Architectural choices documented


Real-World Usage

Planning Framework v2.0 was built using itself (dogfooding):

  • Issue: 20240127-feat-implement-v2
  • 6 phases: Bootstrap, Templates, Scripts, Documentation, Self-Migration, Testing
  • 51 tasks: Tracked in implementation-plan.md
  • Progress: Updated after each session
  • Result: Working v2.0 framework! 🎉

Customization

Issue Types

Default: feat, bug, improve

Add more in PLANNING.md:

- `refactor` - Code restructuring
- `docs` - Documentation only
- `test` - Testing improvements
- `chore` - Maintenance tasks

QA Workflow

Customize .qa-workflow.md:

### Project-Specific
- [ ] Performance: API response < 100ms
- [ ] Security: No SQL injection risks
- [ ] Accessibility: WCAG AA compliant

When to Use

✅ Good Fit

  • Multi-session projects (days/weeks/months)
  • AI-assisted development
  • Multiple feature branches
  • Complex projects with many decisions
  • Team collaboration

⚠️ May Be Overkill For

  • Single-session prototypes
  • Trivial scripts (<100 LOC)
  • Well-defined, simple tasks

What's Different from v1.0?

Feature v1.0 v2.0
Work tracking Global files Issue folders
File growth Unbounded Bounded (roadmap only)
Merge conflicts Yes (planning docs) No (branch-specific)
Multi-agent Separate configs Single PLANNING.md
Issue tracking Manual Built-in workflow
QA gates None .qa-workflow.md

Contributing

Contributions welcome! Areas:

  • ✨ Additional templates
  • 🔧 Platform-specific improvements
  • 📚 Integration examples
  • 🎨 Domain-specific variations
  • 📖 Real-world examples

See CONTRIBUTING.md for guidelines.


Support


License

MIT License - Use freely in any project. See LICENSE for details.


Acknowledgments

Created by: Human + Claude collaboration Inspired by: Need for better context preservation in AI-assisted development Built with: Planning Framework v2.0 (dogfooding!)


Version History

  • v2.0.0 (2024-01-27) - Issue-Based Workflow

    • 🎯 Issue-based workflow (each task in its own folder)
    • 📏 Bounded file sizes (global files stay small)
    • 🚫 No merge conflicts (branch-specific issues)
    • 🤖 Multi-agent support (single PLANNING.md)
    • ✅ Built-in QA workflow
    • 🔧 Automation scripts (setup, migration, helpers)
    • 📖 Comprehensive documentation
    • See CHANGELOG.md for details
  • v1.1.0 (2024-11-12) - Time Estimation Removal

    • Remove time estimation fields
    • Update templates and documentation
  • v1.0.0 (2024-11-05) - Initial Release

    • Core templates (PRD, Implementation Plan, Session Log, Decisions)
    • Setup scripts
    • CLAUDE.md integration

Star this repo if it helps your project! ⭐

Planning Framework v2.0 - Build better with AI assistance 🚀

About

AI planning framework

Resources

License

Contributing

Stars

Watchers

Forks

Packages

No packages published

Contributors 2

  •  
  •