UX Copy Audit & Refactoring System
Turn a CLAUDE.md style guide into an enforced standard across an entire product's content — bulk audits, diff-reviewed refactoring, and zero-risk read-only consistency reports across hundreds of files at once.
Your style guide says 'active voice, no click-here links, sentence case' and half your product still breaks all three rules — because enforcing a style guide by hand across 300 content files means reading all 300 files, and nobody has time for that.
Who it's for: UX writers who currently depend on developers for routine content changes, content designers wanting to enforce a style guide at scale instead of one file at a time, teams doing a pre-launch content audit across hundreds of strings, UX writers hesitant about AI editing production copy who want a zero-risk starting point, content teams tired of inconsistent terminology creeping back in after every review
Example
"Audit our /docs folder for tone violations" → A read-only markdown report flagging every file that breaks the CLAUDE.md style guide, with no files touched — then, once trust is built, a diff-reviewed refactor pass that fixes header capitalization and passive voice across the whole folder, approved change by change rather than blindly applied
CLAUDE.md Template
New here? 3-minute setup guide → | Already set up? Copy the template below.
npx ccplaybooks add ux-copy-audit-system# UX Copy Audit & Refactoring System
## Your Role
You are a content-systems tool for UX writers, not a one-off drafting assistant. Instead of copy-pasting text into a chat window one string at a time, you work directly against the project's actual files — auditing hundreds of strings for consistency, refactoring copy to match house style, and generating read-only reports, all governed by one style guide file you read before touching anything.
If you understand folders, files, and language, that's the whole prerequisite — no engineering background required. The shift this enables: from solving one copy problem at a time to building a repeatable quality process that scales across an entire product's content.
---
## Setup
Get new playbooks like this one
One email a week with new Claude Code workflows. Free, like everything here.
No spam. Unsubscribe anytime.
What This Does
Shifts UX writing from copy-pasting text into a chat window, one string at a time, to running a style guide as an enforced standard across an entire project's files. A single CLAUDE.md file becomes the source of truth for voice, tone, terminology, and formatting — read automatically before every change — and three workflows (bulk audit, diff-reviewed refactor, read-only report) apply it at whatever scale the project actually needs, from one file to hundreds.
Quick Start
Step 1: Create a Project Folder
mkdir ux-copy-audit && cd ux-copy-audit
Step 2: Download the Template
Click Download above, then:
mv ~/Downloads/CLAUDE.md ./
Then edit the style guide sections to match your actual voice, tone, and terminology rules before running anything.
Step 3: Run an Audit
claude
Then ask: "Read all files in [folder] and check for style guide violations — don't edit anything, just report"
The Three Workflows
| Workflow | What it does | Risk level |
|---|---|---|
| Bulk content audit | Finds a pattern (e.g. "click here" links) across every file at once | Read-only, zero risk |
| Automated refactoring | Applies style guide rules directly, shows a diff, waits for Y/N approval per change | Low — every change reviewed |
| Read-only consistency report | Full tone/style check against CLAUDE.md, generates a report, touches nothing |
Zero risk |
Start with the read-only report if the idea of AI editing production copy feels uncomfortable — move to refactoring only once the reports consistently match your own judgment.
Plan Mode vs. Normal Mode
| Plan Mode | Normal Mode | |
|---|---|---|
| Behavior | Proposes, waits for approval | Executes directly |
| Best for | New projects, complex changes | Fast iteration once trust is built |
Always start in Plan Mode.
Tips & Best Practices
- Write a specific style guide, not a vague one. "Be friendly" enforces nothing. "Use 'you' and 'your', avoid passive voice, sentences under 20 words, never 'click here' as link text" is something Claude can actually check against.
- Review every diff before approving a bulk refactor. The value of applying a style guide across 300 files at once is exactly why a single bad automated judgment is worth catching before it propagates — approve change by change, not in one blanket accept.
- Read-only mode is the trust-building step, not a limitation. Ask explicitly for a report with no edits before ever letting the tool touch production copy — this is the fastest way to calibrate confidence in how well your style guide is specified.
- State "do not edit the files" explicitly for audit-only requests. Don't rely on the request's general phrasing to prevent autonomous changes — say it directly.
- This runs on a per-token cost model, not the chat app's flat fee. Typically pennies for a full day of audit work, but worth knowing before pointing it at a very large content folder for the first time.
Limitations
- Data is processed locally but text content is sent to Anthropic's servers — verify your account's data handling settings before running this against confidential or enterprise-sensitive copy
- This is a content-operations and consistency-enforcement tool, not a drafting or ideation tool — for actually writing new copy from scratch, pair it with a UX-writing craft playbook covering voice, patterns, and benchmarks
- Best suited for projects with genuinely large content surfaces (dozens to hundreds of files) — the setup overhead isn't worth it for a five-page site you can review by hand in ten minutes