# Ultrawork

> Parallel execution engine for high-throughput task completion. Activate when user says: ulw, ultrawork, parallel execution, run these in parallel.

- Skill: `zereight/ultrawork` (Agent Skill)
- Install (CLI): `npx skillmds add zereight/ultrawork`
- Raw SKILL.md: https://api.skillmd.com/api/skills/zereight/ultrawork/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: zereight (https://skillmd.com/u/zereight)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/zereight/ultrawork

---


# Ultrawork

Parallel execution engine that runs multiple agents simultaneously for independent tasks. Component mode -- provides parallelism and smart model routing but not persistence or verification loops.

## When to Use
- Multiple independent tasks can run simultaneously
- Need to delegate work to multiple agents at once

## When NOT to Use
- Need guaranteed completion with verification → use `/ralph` (includes ultrawork)
- Full autonomous pipeline → use `/omg-autopilot` (includes ralph which includes ultrawork)
- Only one sequential task → delegate directly to @executor

## Relationship
```
ralph (persistence wrapper)
 └── includes: ultrawork (this skill)

omg-autopilot (autonomous execution)
 └── includes: ralph
     └── includes: ultrawork (this skill)
```

## Steps
1. Classify tasks by independence: parallel vs dependent
2. Fire independent tasks simultaneously via @executor
3. Run dependent tasks sequentially after prerequisites
4. Background long operations (builds, installs, tests)
5. Verify when all tasks complete (build passes, tests pass)

## Execution Policy
- Fire ALL independent tasks simultaneously
- Run builds/tests in background
- Match task complexity to appropriate model routing