refactor-writing

Review and refactor context-bearing writing with evidence-backed findings and a four-page HTML artifact.

Updated Jun 29, 2026
One-click install
npx skills add https://github.com/thorsenk/skills --skill refactor-writing-thorsenk
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: refactor-writing
Source: https://github.com/thorsenk/skills/tree/main/skills/refactor-writing
Command: npx skills add https://github.com/thorsenk/skills --skill refactor-writing-thorsenk

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve? Consequential prose like policies, briefings, handoffs, and instructions often carries ambiguity, repetition, or weak action language, but casual rewriting risks losing protected meaning. This Skill reviews such writing, keeps only changes that earn their context cost, and delivers exact original-versus-proposal comparisons with traceable evidence. ## Core Features & Use Cases - Evidence-backed findings: Each retained recommendation includes the exact original span, proposed replacement, evidence basis, rationale, and preserved meaning, with a real status such as "Recommended — not applied". - No-change branch: If no consequential recommendation survives, it returns "No change recommended" with the unchanged source instead of inventing edits. - Portable HTML artifact: A Node.js renderer generates a self-contained four-page report (report, SKILL.md, method, design system) with three color modes, copy controls, reduced-motion and print fallbacks, and no network dependencies. - Use Case: A team lead asks for a review of an incident handoff document; the Skill identifies three ambiguous instructions, proposes exact replacements with rationale, and delivers a copy-ready revised document plus the full HTML report. ## Quick Start Use the refactor-writing skill to review this policy document, show only consequential evidence-backed changes, and produce the complete HTML artifact.

Frequently Asked Questions about refactor-writing

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

FAQPage Schema
How do I review and rewrite a policy document without losing meaning?▼

Use this Skill to record the document's protected meaning first, then receive only consequential changes with exact original and proposed spans, evidence, and rationale. It ends with one complete copy-ready revision or an explicit no-change outcome.

What does the refactor-writing skill produce as output?▼

It produces a self-contained four-page HTML artifact: a report with verdict and findings, the exact active SKILL.md with its SHA-256 hash, a method page explaining evidence boundaries, and a design-system catalog. All files use same-folder relative links and no network resources.

When should I not use this writing refactor skill?▼

Do not use it for prompt evaluation or hardening, requirements completeness review, source-code refactoring, codebase audits, or casual rewriting that does not need evidence, comparison, and an artifact. It is scoped to consequential prose improvements only.

Does the skill apply changes to my source document automatically?▼

No. It recommends changes by default and only applies them to the reviewed source when the user explicitly authorizes application. Creating the HTML artifact does not count as authorization, and findings are labeled with their real status.

Why does the renderer reject my output folder location?▼

The renderer rejects output folders that are symbolic links or located inside the reviewed source workspace, to keep generated artifacts separate from source material. You can pass --allow-inside-source only after explicit user authorization.