start-work

Execute Prometheus work plans in Codex with Boulder state and evidence tracking.

5|Updated May 17, 2026
One-click install
npx skills add https://github.com/islee23520/lfg --skill start-work-islee23520
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: start-work
Source: https://github.com/islee23520/lfg/tree/main/skills/start-work
Command: npx skills add https://github.com/islee23520/lfg --skill start-work-islee23520

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill manages Prometheus work plans in Codex when you need a disciplined way to select a plan, update Boulder state, and drive execution without losing track of progress.

Core Features & Use Cases

  • Plan selection and bootstrap: Picks the right plan from .omo/plans/ or creates one when none exists.
  • Stateful orchestration: Writes Boulder session state, advances checkboxes, and records evidence in the ledger.
  • Parallel delegation: Spawns subagents for implementation, verification, and review while preserving worktree and cleanup discipline.

Quick Start

Use the start-work skill to execute the selected Prometheus plan from selection through verification, evidence capture, and completion.

Frequently Asked Questions about start-work

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

FAQPage Schema
How do I orchestrate plan execution with subagents in Codex?

Orchestrating plan execution with subagents in Codex is handled by selecting a Prometheus plan, maintaining Boulder state, and spawning parallel subagents for implementation and verification. The skill drives execution from selection through completion while tracking progress.

How does Boulder state track work plan progress across sessions?

Boulder state tracks work plan progress across sessions by writing session state, advancing checkboxes, and recording evidence in the ledger. This mechanism ensures execution continuation and verification gating are preserved when orchestrating tasks across multiple work periods.

What is an evidence ledger in plan-driven implementation workflows?

An evidence ledger in plan-driven implementation workflows is a record that captures verification outputs and task completion proof during orchestrated execution. It maintains cleanup tracking and disciplined orchestration by storing evidence generated by delegated subagents.

Can I spawn parallel subagents for implementation and review tasks?

Spawning parallel subagents for implementation and review tasks is supported through delegated orchestration. The skill coordinates parallel task execution while preserving worktree discipline, verification gating, and cleanup tracking for each subagent's assigned work.

Do I need a pre-existing plan in .omo/plans/ to start work?

A pre-existing plan in .omo/plans/ is not strictly required to start work. The skill picks the right plan from the directory when available, but it can also create a new plan when none exists to bootstrap the orchestration workflow.

When should I not use orchestrated plan execution for subagent coordination?

Orchestrated plan execution for subagent coordination should be avoided when tasks lack clear verification gates or structured work plans. Workflows requiring immediate ad-hoc execution without disciplined cleanup tracking, evidence ledgers, or stateful continuation do not fit this approach.