Shannon

Your AI toolkit to ship faster ⚡

Shannon scaffolds an AI-native product management workspace that works with Claude in Cursor. It gives your AI assistant persistent memory, specialized skills, and structured initiative tracking.

 ███████╗██╗  ██╗ █████╗ ███╗   ██╗███╗   ██╗ ██████╗ ███╗   ██╗
 ██╔════╝██║  ██║██╔══██╗████╗  ██║████╗  ██║██╔═══██╗████╗  ██║
 ███████╗███████║███████║██╔██╗ ██║██╔██╗ ██║██║   ██║██╔██╗ ██║
 ╚════██║██╔══██║██╔══██║██║╚██╗██║██║╚██╗██║██║   ██║██║╚██╗██║
 ███████║██║  ██║██║  ██║██║ ╚████║██║ ╚████║╚██████╔╝██║ ╚████║
 ╚══════╝╚═╝  ╚═╝╚═╝  ╚═╝╚═╝  ╚═══╝╚═╝  ╚═══╝ ╚═════╝ ╚═╝  ╚═══╝

Quick Start

# Create a new workspace with interactive setup
npx create-shannon

# Or specify a directory name
npx create-shannon my-pm-workspace

# Include all skills (not just defaults)
npx create-shannon my-pm-workspace --full

Then open in Cursor and start working with Claude!

What You Get

my-pm-workspace/
├── .claude/
│   ├── CLAUDE.md           # AI memory & PM frameworks
│   ├── CLAUDE.local.md     # Your preferences (gitignored)
│   └── skills/             # Specialized AI capabilities
│       ├── email-drafter/
│       ├── slack-writer/
│       ├── programme-manager/
│       ├── executive-assistant/
│       └── ...more
├── Initiatives/            # Project tracking
│   ├── _template/          # Copy this for new initiatives
│   └── getting-started/    # Tutorial initiative
├── Templates/              # Reusable document templates
├── Operations/             # Recurring deliverables
├── Reference/              # Research & context
└── Output/                 # Generated files (gitignored)

The Memory System

Shannon uses a three-tier memory hierarchy:

| Tier | File | Loaded | Purpose | |------|------|--------|---------| | Project | .claude/CLAUDE.md | Auto | PM frameworks, protocols, skill registry | | Personal | .claude/CLAUDE.local.md | Auto | Your name, role, preferences (gitignored) | | Initiative | Initiatives/*/CLAUDE.md | On-demand | Project-specific decisions, blockers, status |

This means Claude remembers your frameworks every session, knows your preferences, and can pick up initiative context as you work.

Skills

Skills are specialized behaviors that override generic AI responses:

Core Skills (Default)

| Skill | Trigger | What It Does | |-------|---------|--------------| | Email Drafter | "draft an email" | CRAFT method for professional emails | | Slack Writer | "write a slack message" | Voice-matched messaging templates | | Programme Manager | "summarize this meeting" | Comprehensive transcript analysis | | Executive Assistant | "what should I work on" | LNO-prioritized daily TODOs |

Advisory Skills (Optional)

| Skill | Trigger | What It Does | |-------|---------|--------------| | VP Advisor | "VP perspective" | Executive strategic thinking | | Legal Advisor | "legal perspective" | Risk & compliance review | | Sales Advisor | "sales perspective" | Revenue & competitive analysis | | Marketing Advisor | "marketing perspective" | GTM & positioning review | | UX Advisor | "UX review" | Heuristics & accessibility | | Landing Page Evaluator | "evaluate landing page" | 5-dimension scoring |

Execution Skills (Optional, need setup)

| Skill | Trigger | What It Does | |-------|---------|--------------| | Slide Generator | "create slides" | PowerPoint from JSON | | PDF Converter | "convert to PDF" | Markdown to PDF |

How Skills Activate

Skills activate by semantic intent, not exact phrases:

You say: "Help me write an email to the team about the delay"
   ↓
Claude classifies: Communication → Email task
   ↓
Claude reads: .claude/skills/email-drafter/SKILL.md
   ↓
Claude applies: CRAFT method, professional tone, templates
   ↓
Result: Structured, professional email (not generic response)

This is the key difference from vanilla Claude—specialized frameworks consistently applied.

Initiative Tracking

Initiatives are bounded workstreams with their own context:

# Create a new initiative
cp -r Initiatives/_template Initiatives/my-feature-launch

Each initiative's CLAUDE.md tracks:

• Status and timeline
• Key decisions with rationale
• Active blockers and open questions
• Success metrics

Claude auto-updates initiative context as you work!

Daily Workflow

Morning Routine

You: "What should I work on today?"
Claude: [Scans all active initiatives]
        [Extracts blockers, questions, milestones]
        [Creates LNO-prioritized TODO list]

During the Day

You: "Draft an email about pushing the launch to next month"
Claude: [Uses email-drafter skill]
        [Applies CRAFT method]
        [Returns professional, concise email]

After Meetings

You: [Paste transcript] "Summarize this meeting"
Claude: [Uses programme-manager skill]
        [Extracts decisions, action items]
        [Routes updates to relevant initiatives]

Commands

# Interactive setup (recommended for first time)
npx create-shannon

# Quick setup with defaults
npx create-shannon my-project --no-interactive

# Include all 12 skills
npx create-shannon my-project --full

# List all available skills
npx create-shannon skills

# Get help
npx create-shannon --help

Requirements

• Node.js 18+ for the CLI
• Cursor IDE with Claude access
• Python 3.8+ (only for slide-generator and pdf-converter skills)

FAQ

Why not just use Claude directly?

Vanilla Claude forgets your preferences every conversation and gives generic responses. Shannon provides:

• Persistent memory across sessions
• Specialized skills that override generic behavior
• Structured frameworks for consistent quality
• Initiative tracking that accumulates context

Do I need to learn all the skills?

No! Start with the 4 core skills (email, slack, meeting, TODO). Add advisory and execution skills when you need them.

Can I customize the frameworks?

Absolutely. Edit .claude/CLAUDE.md to modify frameworks, add your company context, or adjust protocols.

What if I don't use Cursor?

Shannon is designed for Cursor + Claude, but the .claude/ folder structure works with any AI that reads markdown context files.

Contributing

Found a bug? Have an idea for a new skill?

1. Fork the repository
2. Create a feature branch
3. Submit a pull request

License

MIT

Happy shipping! 🚀