Operate Protocol

uno

AI works FOR you. You define WHAT needs doing. AI handles HOW.

uno is for work that needs delegating but isn't code. The same atomic cycle, lighter scope, focused on getting things done.

When to Use uno

Good Fit

  • • Planning a trip
  • • Tracking a 90-day onboarding
  • • Maintaining documentation
  • • Managing personal projects
  • • Research and synthesis

Not The Right Fit

  • • Building a web app
  • • Writing production code
  • • Automating workflows
  • • Collaborative coding

→ Use duo for code projects

Same Cycle, Lighter Scope

uno uses the same five-phase atomic cycle as every protocol. The difference: narrower context, simpler files.

PROTOCOLLOADCLARIFYEXECUTERECORDREFLECT

Each cycle's RECORD makes the next cycle's LOAD richer.

Which habits matter for uno? Habit 1 (plan in markdown) and Habit 4 (measure compound) are the essentials. You're planning tasks and accumulating context. Habits 2 and 3 (protecting decisions, alignment) matter more in duo where architecture decisions cascade.

LOADReads your project context + persona preferences. Just enough to act.
CLARIFYConfirms what you need. "3 restaurant options near your hotel?"
EXECUTEDoes the work. Plans, researches, drafts, organises.
RECORDWrites what happened. Updates TODO, logs progress.
REFLECTNotes what to remember. "You preferred the coastal route."

Scope Modifiers

uno defaults to narrow scope, just CLAUDE.md + TODO.md + TASKS/. When you need broader context, use a modifier.

CommandScopeWhen to Use
..update [context]TargetedProvide specific context for a focused task
..planFullDesign approach before execution
..reviewFullAudit for inconsistencies

Without a modifier, AI loads narrow scope and runs the atomic cycle. Modifiers expand what LOAD reads. The cycle itself stays the same.

Pick Your Extension

When you start a uno project, AI asks which extension fits your work. Each extension shapes what LOAD reads and what RECORD writes:

uno/ea/Executive Assistant

Work projects with deadlines, stakeholders, learning curves

Creates: UPDATES.md, GLOSSARY.md, KNOWLEDGE.md, SYSTEMS.md

Example: 90-day onboarding, project tracking, stakeholder management

uno/pa/Personal Assistant

Personal logistics. Trips, events, moves, finances.

Creates: ITINERARY.md, BOOKINGS.md, BUDGET.md, NOTES.md

Example: Trip planning, event coordination, personal projects

uno/km/Knowledge Manager

Documentation, archives, recurring tasks

Creates: LOG.md, PLANNED.md

Example: Team knowledge base, documentation maintenance

Personalisation Layer

After picking your extension, AI runs a CLARIFY phase to learn how you think. This creates PERSONA.md, loaded on every future interaction so AI already knows your preferences.

AspectWhat AI LearnsHow It's Used
Decision styleHow you prefer choices presentedStructures recommendations
Signal watchWhat to always surfaceFilters briefings
Noise filterWhat to skip or minimiseReduces clutter
Interests/constraintsWhat matters to youShapes suggestions

Example: PA Persona for Trip Planning

## Priority Signals
1. Local/authentic experience
2. Value for money
3. Avoiding crowds

## Constraints
- Vegetarian options required
- Nothing before 9am
- Budget: $150-250/night

## Interests
Local food markets, street art, live music, coastal walks

## Avoid
Chain restaurants, tourist traps, hop-on hop-off tours

This is written to PERSONA.md during onboarding. Every future LOAD reads it. AI never asks "what kind of food do you like?" twice.

Core Files

Every uno project has these:

my-project/
├── CLAUDE.md      # Protocol + project reference (ground truth)
├── PERSONA.md     # Your preferences and signals
├── TODO.md        # Tasks with acceptance criteria
├── PROGRESS.md    # Session log
├── TASKS/         # Human-only tasks
├── README.md      # Project description
└── _archive/      # Old entries

Plus extension-specific files based on your choice (ea, pa, km, or none).

Task Ownership

Location encodes ownership, no markers needed.

LocationOwnerPurpose
TODO.mdAIClaude's work queue
TASKS/HumanYour work queue (TODO, PRIORITY, BLOCKED, DONE)

If it's in TODO.md, AI owns it. If it's in TASKS/, you own it. No ambiguity.

Quick Start

1. Clone the template

git clone https://github.com/marekgorski/protocol-uno.git my-project
cd my-project
rm -rf .git && git init

2. Open with Claude

Claude reads CLAUDE.md (LOAD), then runs onboarding (CLARIFY):

  • • Extension selection (ea, pa, km, none)
  • • 5 project questions (what, who, problem, success, constraints)
  • • Persona questions based on extension
  • • File population (EXECUTE) and commit (RECORD)

3. Start working

"Plan my trip to Lisbon"  # LOAD → CLARIFY → EXECUTE → RECORD → REFLECT
"Book the flights"        # LOAD → CLARIFY → EXECUTE → RECORD → REFLECT

# Every interaction runs the full cycle. Context accumulates automatically.

Ready to Get Started?

Clone the template and let AI guide you through setup.