Shannon Framework
Version 5.6.0 | Installation | Quick Start | Documentation | GitHub
🚀 V5.6.0 "Comprehensive Quality & Intelligence": 14 new skills, /shannon:health dashboard, advanced testing, security automation!
See docs/v5.6/CHANGELOG.md for complete details.
Overview
Shannon Framework is a Claude Code plugin that transforms AI-assisted development from ad-hoc interactions into a rigorous, quantitative, reproducible methodology. Built for mission-critical domains—Finance, Healthcare, Legal, Security, Aerospace—Shannon replaces subjective judgments with objective measurements throughout the development lifecycle.
What Shannon IS
- Complete Development Methodology: Specification → Planning → Execution → Testing → Reflection
- Multi-Layer Enforcement Architecture: 4 layers (Commands → Skills → Hooks → Core Files) enforcing same principles
- Quantitative Framework: Replace "seems complex" with 8D complexity scores (0.00-1.00)
- Production-Tested System: Built by observing actual LLM violations, creating explicit counters
What Shannon IS NOT
- ❌ Code generation tool
- ❌ Testing framework
- ❌ Project scaffolder
- ✅ Complete methodology for specification-driven development with quantitative rigor
Core Value Proposition
Quantitative Over Qualitative: Replace ALL subjective judgments with objective measurements.
| Traditional Approach | Shannon Framework |
|---|
| "Seems complex" | 8D complexity score: 0.72/1.00 (VERY COMPLEX) |
| "Need 3-4 agents" | Formula: complexity × bands = 8-15 agents |
| "Making progress" | Quantified: 65% complete (+12% this session) |
| "Tests look fine" | NO MOCKS compliance: 100% or fail |
| "Architecture aligns" | Alignment score: 87/100 (STRONG) |
Architecture
4-Layer Enforcement Pyramid
USER
│
↓
┌──────────────────────────┐
│ COMMANDS (Layer 4) │ ← User-facing interface
│ 15 slash commands │ /shannon:spec, /shannon:wave
└────────────┬─────────────┘ Delegate to skills
↓ Format output
┌──────────────────────────┐
│ SKILLS (Layer 3) │ ← Workflow implementation
│ 18 behavioral patterns │ Execute methodologies
└────────────┬─────────────┘ Save to Serena
↓ Reference core files
┌──────────────────────────┐
│ HOOKS (Layer 2) │ ← Automatic enforcement
│ 5 lifecycle scripts │ Block violations
└────────────┬─────────────┘ Inject context
↓ Cannot be skipped
┌──────────────────────────┐
│ CORE FILES (Layer 1) │ ← Foundational specs
│ 9 reference documents │ 11K lines total
└──────────────────────────┘ Complete algorithms
↓
SERENA MCP
(Central State Store)
Component Inventory
| Layer | Count | Purpose | Examples |
|---|
| Commands | 20 | User entry points | shannon:init, shannon:do, shannon:wave, shannon:spec, shannon:ultrathink |
| Skills | 21 | Workflow implementation | project-onboarding, spec-analysis, wave-orchestration, intelligent-do |
| Hooks | 5 | Automatic enforcement | post_tool_use (NO MOCKS), precompact |
| Core Files | 9 | Reference specs (11K lines) | TESTING_PHILOSOPHY, SPEC_ANALYSIS |
| Agent Guides | 25 | Agent usage documentation | WAVE_COORDINATOR, TEST_GUARDIAN |
Infrastructure Dependencies
Mandatory:
- Claude Code: Plugin platform (required)
- Serena MCP: Persistent state store (61% of skills require it)
Recommended:
- Sequential MCP: Deep reasoning (100+ thoughts for complex analysis), REQUIRED for
/shannon:ultrathink
- Context7 MCP: Framework-specific documentation
- Tavily MCP: Best practices research
Conditional (based on project type):
- Puppeteer MCP: Web testing (NO MOCKS requirement)
- iOS Simulator MCP: Mobile testing
- Platform-specific MCPs: As needed for real system testing
Installation
Prerequisites
- Claude Code (latest version)
- Serena MCP (mandatory - 61% of skills require it)
- Sequential MCP (recommended for complex analysis)
Installation Steps
⭐ Recommended: Universal Installation (Works for both Claude Code and Cursor IDE)
# Clone repository
git clone https://github.com/shannon-framework/shannon.git
cd shannon
# Install for both editors
./install_universal.sh
krzemienski