ultrawork

Execute independent tasks in parallel with acceptance criteria and evidence collection.

2|Updated May 11, 2026
One-click install
npx skills add https://github.com/xz1220/oh-my-kimi --skill ultrawork-xz1220
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ultrawork
Source: https://github.com/xz1220/oh-my-kimi/tree/main/skills/ultrawork
Command: npx skills add https://github.com/xz1220/oh-my-kimi --skill ultrawork-xz1220

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It solves the time waste of running independent tasks serially by providing a controlled parallel execution engine that still produces lightweight evidence for safe wrap-up.

Core Features & Use Cases

  • High-throughput parallel execution: Runs multiple independent tasks concurrently to increase completion speed.
  • Execution discipline with explicit acceptance criteria: Requires pass/fail verification standards before editing or delegating, reducing “hope-based” completion.
  • Dual-channel approach (direct execution + evidence lanes): Supports a direct tool lane for immediate work and optional background evidence lanes for tests, scoped research, and regression mapping.

Quick Start

Use ultrawork to run multiple independent subtasks in parallel while you define clear acceptance criteria and collect evidence to confirm completion, for example ask it to update two separate modules and run the relevant build and tests to verify the changes.

Frequently Asked Questions about ultrawork

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I execute independent development tasks in parallel safely?

Parallel task execution uses a controlled engine to run independent subtasks concurrently, enforcing explicit pass/fail acceptance criteria to verify completion and collect lightweight evidence.

What is lightweight evidence collection in concurrent workflows?

Lightweight evidence collection operates via optional background lanes to gather scoped testing and research data during concurrent execution, confirming task completion without heavy architect validation overhead.

How do I set acceptance criteria for parallel task delegation?

Define explicit pass/fail verification standards for each subtask before delegating, requiring the parallel engine to meet these acceptance criteria and collect evidence prior to marking the work complete.

Does parallel task execution work for scoped verification workflows?

Yes, the dual-channel approach supports scoped verification workflows by using a direct tool lane for immediate implementation and optional background evidence lanes for concurrent regression mapping and testing.

When should I avoid high-throughput parallel execution for subtasks?

Avoid high-throughput parallel execution for tasks requiring Ralph-level persistence, architect validation, or verified persistence guarantees, as this engine uses lightweight validation and explicit stop conditions instead.