ulw-execute

Orchestrate autonomous agent workflows for Prometheus work plans using parallel subagent delegation and verification.

68.6k|5.6k|Updated Dec 3, 2025
One-click install
npx skills add https://github.com/code-yeongyu/oh-my-opencode --skill ulw-execute
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ulw-execute
Source: https://github.com/code-yeongyu/oh-my-opencode/tree/main/packages/shared-skills/skills/ulw-execute
Command: npx skills add https://github.com/code-yeongyu/oh-my-opencode --skill ulw-execute

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill solves the challenge of managing complex, multi-step software development tasks by enforcing strict orchestration, parallel delegation, and rigorous verification, preventing the common pitfalls of manual implementation and unverified code changes.

Core Features & Use Cases

  • Autonomous Orchestration: Automatically decomposes Prometheus plans into atomic tasks and dispatches them to specialized subagents.
  • Rigorous Verification: Enforces a five-gate verification process, including automated tests, manual QA, and adversarial probing for every task.
  • Use Case: When tasked with a large refactor, this Skill spawns parallel workers to handle independent modules, monitors their progress, verifies their output against characterization tests, and manages the final PR lifecycle.

Quick Start

Execute the plan named feature-x-refactor by running the ulw-execute command with the appropriate worktree path.

Frequently Asked Questions about ulw-execute

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

FAQPage Schema
How do I execute a Prometheus plan with ulw-execute?

Run the ulw-execute command followed by the plan name and the required worktree path. The skill will then bootstrap the Boulder state and begin dispatching tasks to subagents.

What is the role of the orchestrator in this skill?

The orchestrator acts as a manager that delegates all implementation, testing, and QA to spawned subagents. It never writes code or edits product files directly, ensuring strict separation of concerns.

How does the skill handle task verification?

Every task must pass five gates: plan reread, automated verification, manual QA, adversarial QA, and cleanup. Only after these gates are cleared is a task marked as fully done in the ledger.

Can I use this skill for PR-based development?

Yes, use the --make-pr or --ship flags to automate the creation of task-owned git worktrees, PR generation, and the full delivery lifecycle including CI monitoring and merging.

What happens if a subagent fails a task?

The orchestrator records the failure in the ledger, resets the checkbox to in-progress, and re-dispatches the executor with the specific failure feedback for correction.