Beastmode × Spec-Kit Synthesis: The definitive project creation and management system combining disciplined engineering workflow with specification-driven development.
┌─────────────────────────────────────────────────────────────────────────────┐
│ │
│ SPECIFY → PLAN → TASKS → IMPLEMENT → VALIDATE → RELEASE │
│ │ │ │ │ │ │ │
│ ▼ ▼ ▼ ▼ ▼ ▼ │
│ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ │
│ │ Spec │ │ Plan │ │Tasks │ │ Code │ │ Test │ │ Ship │ │
│ │ .md │ │ .md │ │ .md │ │ │ │ │ │ │ │
│ └──────┘ └──────┘ └──────┘ └──────┘ └──────┘ └──────┘ │
│ │
│ With: Constitution • Knowledge Hierarchy • Swarm Mode • Retro Loop │
│ │
└─────────────────────────────────────────────────────────────────────────────┘
The Ultimate Project System unifies two powerful methodologies:
| Beastmode | Spec-Kit | Ultimate System |
|---|---|---|
| Disciplined 5-phase workflow | Spec-Driven Development | Executable specs with workflow discipline |
| L0-L3 knowledge hierarchy | Feature specifications | Persistent specs that compound knowledge |
| Retro learning loop | Constitutional governance | Self-improving system with immutable principles |
| Swarm mode (parallel agents) | Research agents | Parallel execution with research context |
| Progressive autonomy gates | Phase gates | Progressive autonomy with constitutional enforcement |
A system where:
- ✅ Specifications are executable - they generate implementation
- ✅ Knowledge persists - across sessions, features, and team members
- ✅ Architecture is governed - by immutable constitutional principles
- ✅ Workflow is disciplined - no skipping phases, retro after each
- ✅ Execution is efficient - parallel swarm mode where safe
- ✅ Projects self-improve - retro loop captures and promotes learnings
# 1. Clone this repository
# This is a synthesis of two existing projects:
# - Beastmode: https://github.com/BugRoger/beastmode
# - Spec-Kit: https://github.com/github/spec-kit
#
# To use this system, you can either:
# 1. Copy the files directly from this repository
# 2. Create your own synthesis based on the upstream projects
#
# For the original projects, visit:
git clone https://github.com/github/spec-kit.git
# or
git clone https://github.com/BugRoger/beastmode.git
cd ultimate-project
# 2. Run the setup script for your AI agent
./scripts/init.sh --agent kimi --project ../my-new-project
# 3. Follow the guided setup
cd ../my-new-project
# Then tell your AI: /constitution establish# 1. Clone this repository
git clone https://github.com/yourusername/ultimate-project.git
cd ultimate-project
# 2. Run the migration script
./scripts/init.sh --agent kimi --project ../my-existing-project --existing
# 3. Review and commit the changes
cd ../my-existing-project
git add .
git commit -m "chore: Add Ultimate Project System"See docs/INSTALLATION.md for detailed manual setup instructions.
ultimate-project/ # This repository
├── README.md # This file
├── docs/ # Comprehensive documentation
│ ├── ARCHITECTURE.md # System architecture deep dive
│ ├── COMMANDS.md # Command reference
│ ├── INSTALLATION.md # Setup for new/existing projects
│ ├── CONSTITUTION.md # The Nine Articles explained
│ ├── SWARM_MODE.md # Parallel execution guide
│ └── VISUAL_GUIDE.md # Diagrams and flowcharts
├── templates/ # Specification templates
│ ├── spec-template.md # Feature specification
│ ├── plan-template.md # Implementation plan
│ └── tasks-template.md # Task list
├── scripts/ # Setup and utility scripts
│ ├── init.sh # Project initialization
│ ├── install-kimi.sh # Kimi-specific setup
│ └── verify.sh # Verify installation
├── examples/ # Example projects
│ ├── new-project/ # Greenfield example
│ └── existing-project/ # Brownfield example
└── src/ # System source (if needed)
When you initialize a project, it gets this structure:
your-project/
├── .beastmode/ # Beastmode: Knowledge & workflow
│ ├── BEASTMODE.md # L0: System manual
│ ├── config.yaml # Configuration
│ ├── context/ # L1-L2: Published knowledge
│ ├── meta/ # L1-L2: Learnings & SOPs
│ ├── state/ # L3: Checkpoint artifacts
│ └── worktrees/ # Git worktrees
├── memory/ # Spec-Kit: Project memory
│ ├── constitution.md # The Nine Articles
│ └── guidance.md # Runtime guidance
├── .beastmode/specs/ # Spec-Kit: Feature specs
│ └── NNN-feature/ # Numbered features
│ ├── spec.md
│ ├── plan.md
│ ├── tasks.md
│ └── ...
└── src/ # Implementation
Immutable architectural principles:
- Library-First - Every feature begins as a standalone library
- CLI Interface - All libraries expose functionality via CLI
- Test-First - TDD mandatory - tests before implementation
- Integration Testing - Real dependencies over mocks
- Observability - Everything inspectable through text I/O
- Versioning - Semantic versioning
- Simplicity - ≤3 projects, YAGNI principles
- Anti-Abstraction - Use frameworks directly
- Integration-First - Contract tests before implementation
| Level | Path | Content | When Loaded |
|---|---|---|---|
| L0 | .beastmode/BEASTMODE.md |
System manual, constitution | Autoload every session |
| L1 | .beastmode/context/{PHASE}.md |
Phase summaries | At phase start |
| L2 | .beastmode/context/{phase}/{domain}.md |
Detail docs | On-demand |
| L3 | .beastmode/specs/NNN-*/, .beastmode/state/ |
Specs, artifacts | Checkpoint |
specify → plan → tasks → implement → validate → release
│ │ │ │ │ │
└────────┴───────┴─────────┴──────────┴─────────┘
│
[RETRO after each]
│
▼
Knowledge promotion (L3→L2→L1→L0)
Each phase: prime → execute → validate → checkpoint → retro
Parallel execution using AI agent Task spawning:
# Research multiple topics in parallel
research_tasks = [
Task(agent="researcher", topic="JWT auth"),
Task(agent="researcher", topic="Session cookies"),
Task(agent="researcher", topic="OAuth2")
]
findings = [t.result() for t in research_tasks]
# Execute parallel-safe waves
if wave.parallel_safe and no_file_overlap(tasks):
agents = [Task(agent="implementer", task=t) for t in tasks]
results = [a.result() for a in agents]
# Run retro agents simultaneously
context_task = Task(agent="retro-context", ...)
meta_task = Task(agent="retro-meta", ...)
results = [context_task.result(), meta_task.result()]| Document | Description |
|---|---|
| docs/ARCHITECTURE.md | Deep dive into unified architecture |
| docs/INSTALLATION.md | Setup for new and existing projects |
| docs/COMMANDS.md | Complete command reference |
| docs/CONSTITUTION.md | The Nine Articles explained |
| docs/SWARM_MODE.md | Parallel execution patterns |
| docs/VISUAL_GUIDE.md | Diagrams and flowcharts |
| Command | Description |
|---|---|
/constitution [principles] |
Establish project constitution |
/specify [description] |
Create feature specification |
/research [topics] |
Research technical unknowns |
| Command | Description |
|---|---|
/plan [tech-stack] |
Create implementation plan |
/tasks |
Generate task list |
| Command | Description |
|---|---|
/implement [--parallel] |
Execute implementation |
/validate |
Run validation |
/release |
Release feature |
/status |
Show project status |
| Aspect | Traditional | Ultimate System |
|---|---|---|
| Spec to Code | Manual translation | Executable generation |
| Knowledge | Lost between sessions | Persistent L0-L3 hierarchy |
| Architecture | Ad-hoc decisions | Constitution-governed |
| Parallel Work | Risky, uncoordinated | Safe swarm mode |
| Learning | Individual, lost | Captured in retro, promoted |
| Time to MVP | Days | Hours |
Building a User Authentication System:
Traditional Approach:
├─ Write PRD (4 hours)
├─ Technical design (4 hours)
├─ Implementation (16 hours)
├─ Testing (8 hours)
└─ Total: 4 days
Ultimate System:
├─ /specify (5 min)
├─ /plan (10 min)
├─ /tasks (2 min)
├─ /implement --parallel (45 min)
├─ /validate (5 min)
└─ Total: 72 minutes
With better quality (constitution-enforced, test-first, documented).
| Agent | Support | Notes |
|---|---|---|
| Kimi Code CLI | ✅ Full | Native Task tool support |
| Claude Code | ✅ Full | Via beastmode-claude bridge |
| GitHub Copilot | ✅ Full | Via VS Code extension |
| Cursor | ✅ Full | Via cursor rules |
| Windsurf | ✅ Full | Via workflows |
| Generic | ⚡ Partial | Bring your own adapter |
- Git 2.30+
- Python 3.11+ (for utilities)
- An AI agent (Kimi, Claude, Copilot, etc.)
git clone https://github.com/yourusername/ultimate-project.git
cd ultimate-project
./scripts/init.sh --agent kimi --project ../my-projectSee docs/INSTALLATION.md for detailed options.
- Read docs/CONSTITUTION.md - Understand the Nine Articles
- Try the examples/new-project/ example
- Run through a simple feature end-to-end
- Read docs/ARCHITECTURE.md - Understand the synthesis
- Customize templates in
templates/ - Configure gates in
.beastmode/config.yaml - Experiment with swarm mode
- Read docs/SWARM_MODE.md - Master parallel execution
- Implement custom retro agents
- Extend the constitution for your domain
- Contribute back to the project
Apache 2.0 - See LICENSE
- Beastmode: https://github.com/BugRoger/beastmode - Disciplined engineering workflow system
- Spec-Kit: https://github.com/github/spec-kit - Spec-Driven Development toolkit
- Contributors: The community that shaped both systems
- 📖 Documentation
- 🐛 Issues
- 💬 Discussions
Version: 1.0.0
Last Updated: 2026-03-08
Status: Production Ready
"The ultimate project creation system: disciplined yet flexible, prescriptive yet learning, structured yet efficient."