refine

Run iterative review-driven edit cycles until drafts meet target quality scores.

37|7|Updated Mar 19, 2026
One-click install
npx skills add https://github.com/unclejobs-ai/second-claude-code --skill refine-unclejobs-ai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: refine
Source: https://github.com/unclejobs-ai/second-claude-code/tree/main/skills/refine
Command: npx skills add https://github.com/unclejobs-ai/second-claude-code --skill refine-unclejobs-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It fixes drafts that fail review by running repeatable review-and-edit cycles until the content meets a defined quality target without losing the author’s original intent.

Core Features & Use Cases

  • Review-Driven Refinement Loop: dispatches real subagent reviews via /second-claude-code:review, then applies changes based on ranked feedback rather than guesswork.
  • DoD-Gated Iterations (Optional): when --dod is provided, it evaluates each criterion independently, prioritizes failing criteria, and only stops when all DoD items pass.
  • Safe, Resumable Editing: stores baseline hash/content and maintains an on-disk iteration state (refine-active.json), using git-aware revert safeguards to prevent overwriting unrelated user changes.

Quick Start

Use the refine skill to improve an existing draft file you provide and stop only after the final quick review reports approval or minor fixes meeting your DoD target.

Frequently Asked Questions about refine

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

FAQPage Schema
How do I iteratively edit a draft until it passes quality checks?

To iteratively edit a draft until it passes quality checks, you need a review-driven refinement loop that dispatches LLM reviews, applies ranked feedback, and stops only when a target verdict or score is met.

Can I define a criteria checklist to gate draft refinement?

Yes, you can gate draft refinement by providing an optional Definition of Done checklist, which evaluates each criterion independently and stops the edit cycles only when all DoD items pass consensus.

How does LLM review work for automated draft refinement?

LLM review for automated draft refinement works by dispatching a subagent to evaluate the content, ranking the feedback, and applying changes in repeatable cycles until the output reaches the defined quality target.

Is there a safe way to revert iterative content edits without losing other changes?

Yes, safe revert logic uses baseline hashing and git-aware rollback safeguards to restore content, preventing the overwriting of unrelated user changes during the iterative editing process.

What is the best way to fix an initial output that is close but fails review?

The best way to fix an initial output that fails review is running repeatable review-and-edit cycles that apply ranked feedback until the content meets your defined success criteria without losing original intent.