The Stella Protocol

A structured way to build products with AI — from idea to launch.

Stella Protocol gives your AI coding agent (Claude Code, Cursor, etc.) a set of phases, quality gates, and scope rules so you stay in control of what gets built. You make the decisions. The AI executes with guardrails.

Quick Start

npx stella-protocol install

That's it. This installs:

• 12 Agent Skills — 5 phase orchestrators + 6 atomic skills (governance + execution rigor) + 1 meta skill
• A brain/ directory — markdown files that track your project's decisions, scope, and state

No config needed. Start talking to your AI agent about what you want to build.

Your First Project

Step 1: Brainstorm

Tell your AI agent what you're thinking about:

"I'm thinking about building a community riddle platform"

The IMU satellite activates automatically. It helps you explore the idea, identify risks, and decide whether to go lightweight (East Blue) or full process (Grand Line).

Step 2: Define (Grand Line) or Skip to Build (East Blue)

• Grand Line — For bigger projects. Creates a full PRD with requirements, architecture, UX design. The Shaka satellite walks you through it.
• East Blue — For small features. You get a Mini-PRD and jump straight to building.

Step 3: Build

Tell your agent to start building. The Edison satellite writes the code, pauses for review after significant features, and tracks every decision. If something drifts from your approved scope, Cipher Pol flags it automatically.

Step 4: Review

Before anything goes live, Lilith Red runs a security audit and Lilith Blue checks quality. Critical issues block deployment — you decide what to fix now vs. later.

Step 5: Close

York syncs documentation with reality and Morgans helps craft your launch announcement.

How It Works

Phases

IDEATE  →  Brainstorm, explore, decide if it's worth building
DEFINE  →  Requirements, architecture, UX, PRD
BUILD   →  Implementation with scope monitoring and review checkpoints
REVIEW  →  Security audit, QA, quality gates
CLOSE   →  Documentation, launch, go-to-market

Governance (Always On)

Cipher Pol — Scope enforcement. Detects when AI adds features outside your approved spec. You decide to approve, reject, or amend.

Buster Call — Quality veto. Any satellite can halt progress on critical security or quality issues. You decide how to resolve it.

Project Brain (brain/ directory)

Every decision, every scope change, every status update lives in brain/ as plain markdown — committed to git, reviewable in PRs:

brain/
├── log-pose.md        # Current phase, track, blockers, active work
├── vivre-cards.md     # Append-only decision log
├── scope-changes.md   # Scope drift audit trail
├── architecture.md    # Tech decisions with rationale
├── design-system.md   # Visual design tokens and rules
├── ideas.md           # Idea backlog
└── prd-*.md           # PRD documents (created as needed)

Two Tracks

Satellites

Satellites are specialized AI behaviors that activate from conversational context. You don't invoke them — you talk naturally and the right expertise surfaces:

Agent Skills

Skills are organized in layers. Phase orchestrators handle user-facing flow; atomic skills handle governance and execution rigor — they activate when their trigger conditions match, keeping per-invocation token load low.

Phase orchestrators (5):

Edison atomic (execution rigor, v0.6.0+):

Governance atomic (cross-phase, v0.6.0+):

Meta (skill authoring, v0.7.1+):

Compatible with Claude Code, Cursor, and any tool supporting the Agent Skills open standard.

Example

See the full walkthrough — building House of Riddle, a community riddle platform, through the entire Stella cycle with populated brain/ files.

Philosophy

You are not "talking to an AI agent." You are thinking, and specialized facets of your extended cognition activate to help you execute.

The governance protocols aren't tools you invoke — they're constitutional rules that every satellite enforces. The protocol in protocol/ is human-readable markdown. You can apply it manually without any tooling.

License

MIT

Author

Aditya Uttama — Product Manager who builds 100% with AI.

• LinkedIn
• GitHub