work-loop

Impose a plan-execute-gate-review loop on non-trivial coding work.

15|5|Updated May 4, 2026
One-click install
npx skills add https://github.com/eugenelim/agent-ready-repo --skill work-loop
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: work-loop
Source: https://github.com/eugenelim/agent-ready-repo/tree/main/.agents/skills/work-loop
Command: npx skills add https://github.com/eugenelim/agent-ready-repo --skill work-loop

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill prevents non-trivial coding work from drifting into guesswork by forcing a disciplined loop of planning, execution, verification, and review. It replaces vague progress with explicit gates, recorded state, and clear termination criteria so agents do not declare victory too early.

Core Features & Use Cases

  • Loop discipline: It guides substantial changes through plan, execute, gate, and review phases instead of ad hoc iteration.
  • Mechanical validation: It supports objective checks such as tests, linting, type checks, and other pass-fail gates.
  • Stateful workflow control: It uses bundled scripts and session state to track approval, iterations, findings, and worktree progress.
  • Use case: A developer making a multi-file refactor can use this Skill to write the plan first, run the right validations, review the diff adversarially, and only then mark the work complete.

Quick Start

Use the work-loop skill to plan, execute, verify, and review your non-trivial change, then report any blockers or completed gates.

Frequently Asked Questions about work-loop

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

FAQPage Schema
How do I keep coding agents from declaring victory too early on complex refactors?

To keep coding agents from declaring victory too early, apply a disciplined plan-execute-gate-review loop that requires persisted specs, deterministic validation gates, and adversarial review before marking work complete.

What is the best way to manage state and approval for multi-file coding tasks?

State management for multi-file coding tasks requires tracking plan approval, iterations, and worktree progress using session state, ensuring explicit gates and recorded state replace vague progress updates.

How does an adversarial review loop work for infrastructure or build-system edits?

An adversarial review loop for infrastructure edits examines the diff critically, enforces deterministic pass-fail gates like linting and type checks, and follows recovery paths if iterations stall or fail.

When do I need upfront tests and persisted specs for a coding workflow?

Upfront tests and persisted specs are needed for non-trivial coding workflows like migrations, dependency upgrades, and schema changes to prevent drifting into guesswork and ensure clear termination criteria.

Does this plan-execute-gate-review approach work for dependency upgrades and migrations?

Yes, the plan-execute-gate-review approach works for dependency upgrades and migrations by imposing mechanical validation, stateful workflow control, and clear termination criteria on substantial changes.

Why does my agent drift into guesswork during multi-file bug fixes?

Agents drift into guesswork during multi-file bug fixes because vague progress replaces explicit gates; enforcing a loop of planning, execution, verification, and review prevents this drift.