Key Features

Features born from necessity. We built these because existing tools weren't good enough.

⚡ Ultrawork

The "God Mode" command. Type ultrawork (or ulw) and the system activates. It doesn't just run one prompt; it spins up the entire agent team to tackle the task until it's done.

omo ultrawork 'Refactor the entire auth module to use Clerk'

🔒 Hash-Anchored Edits

The Harness Problem: Most agents fail because they try to edit code by guessing line numbers or rewriting context. If the file changes in the background, the agent corrupts the file.

The Solution: We implemented Hashline. Every line the agent reads is tagged with a content hash. When the agent submits an edit, it must reference these hashes. If the hash doesn't match (because the line changed), the edit is rejected before corruption occurs.

Example of Hash-Tagged Output

11#VK: function hello() {
22#XJ:   return "world";
33#MB: }

🔁 Ralph Loop

A self-referential loop that prevents the agent from stopping prematurely. Instead of asking "Continue?", the agent checks its own completion criteria. If the task isn't 100% done, it loops back into execution mode automatically.

👮 Todo Enforcer

Agents get lazy. They hallucinate completion. The Todo Enforcer monitors the agent's state. If it detects idleness or premature termination on an incomplete task, it yanks the agent back into the workflow.

📂 Deep Initialization

Run /init-deep to auto-generate hierarchical AGENTS.md files throughout your project. This gives agents context-aware instructions specific to each directory, keeping the global context window lean.

project/
├── AGENTS.md              ← Global context
├── src/
│   ├── AGENTS.md          ← Source-specific rules
│   └── components/
│       └── AGENTS.md      ← UI component guidelines