Neo Protocol

Load expert mental models on-demand to enhance reasoning.

Commands

| Command | Action | |---------|--------| | neo | Show Crew status (loaded ✓ vs unloaded ○) | | neo help | List all commands | | neo <module> | Load module. If from Library → auto-add to Crew | | neo <module> off | Unload specific module (stays in Crew) | | neo off | Unload ALL modules | | neo browse | Browse full Library by category | | neo search <query> | Search Library for modules | | neo add <module> | Add to Crew without loading | | neo remove <module> | Remove from Crew (back to Library) | | neo create <description> | Create new module for Library | | neo delete <module> | Delete module from Library/Crew permanently |

Workflow

On neo (no args)

Display Crew status with help hint:

🧠 Neo Protocol

LOADED:
✓ psychologist
✓ game-theorist

CREW:
○ negotiator
○ entrepreneur

RECENTS:
○ dermatologist
○ cosmetic-plastic-surgeon

neo help for commands

Recents shows recently used modules that aren't in Crew. Track in assets/recents.json with timestamps. Persists across sessions. Cap at 5 (oldest drops off). Expires after 1 week of no use.

On neo help

List all available commands.

On neo <module>

1. Find module in Crew or Library
2. Read the module file into context
3. If from Library (not Crew), add to Crew
4. Confirm: "🧠 <module> loaded. [summary of mindset]"

On neo <module> off

1. Note module is unloaded (remove from active context tracking)
2. Module stays in Crew for easy reload
3. Confirm: "🧠 <module> unloaded."

On neo off

1. Clear all loaded modules
2. Confirm: "🧠 All modules unloaded."

On neo create <description>

1. Parse the description for expertise type
2. Generate module using TEMPLATE.md structure
3. Save to assets/library/<category>/<name>.md
4. Add to registry.json
5. Confirm and offer to load

Files

• scripts/neo.py — CLI for library management
• references/TEMPLATE.md — Module creation template
• assets/crew.json — User's personal Crew (gitignored)
• assets/crew.default.json — Starter Crew (ships with skill)
• assets/registry.json — Library index with descriptions
• assets/library/ — All expertise modules by category

First Run

If crew.json doesn't exist, copy crew.default.json → crew.json to initialize the user's personal Crew.

Module Structure

Each module follows this structure:

• Core Mindset — 4-5 key mental traits
• Framework — 4-step systematic approach
• Red Flags — 6 warning signs (🚩)
• Key Questions — 5 essential questions
• Vocabulary — 5 domain terms
• When to Apply — 4 trigger situations
• Adaptations Log — User customizations

State Management

Track loaded modules in conversation context. When user says "neo off" or session ends, consider all modules unloaded. Crew persists in crew.json.

Updates & Customization

Modules in registry.json have a source field:

• "upstream" — Came with the skill, updated by ClawHub
• "custom" — User-created, never touched by updates

And a deleted field for upstream modules:

• false — Active, will be updated
• true — User removed, won't be restored on update

Update behavior:

| Source | Deleted | On Update | |--------|---------|-----------| | upstream | false | ✅ Update normally | | upstream | true | ⏭️ Skip (user removed it) | | custom | — | 👤 Never touched |

Update script:

# Check status
python3 scripts/update.py status

# Merge upstream updates
python3 scripts/update.py merge --upstream /path/to/new/neo

# Delete a module (marks upstream as deleted, removes custom)
python3 scripts/update.py delete --module physicist

# Restore a deleted upstream module
python3 scripts/update.py restore --module physicist

On neo delete <module>:

• If upstream: Set deleted: true (can be restored)
• If custom: Actually remove from registry and library