Purpose
Audit a batch of copy against a brand's defined voice and flag exactly where it drifts off-brand — with line-level rewrites, per-piece scores, and a diagnosis of the systemic patterns causing inconsistency — so every channel and market sounds like the same brand.
When to Use
Invoke this skill when the user asks to:
- Check whether copy matches their brand voice or tone guidelines
- Audit content across platforms (web, email, social, ads, product UI) for consistency
- Compare copy across markets or regions for voice alignment
- Catch voice drift after adding new writers, agencies, or channels
- Onboard a new writer or freelancer by showing where their draft misses the brand voice
- Review a content batch before publishing for on-brand consistency
Inputs Required
Before starting, collect:
- Copy samples: The pieces to audit — paste text, ideally labeled by channel or market
- Brand voice definition: Voice attributes, tone guidelines, or a style guide (paste what exists)
- Voice do's and don'ts: Words/phrases to use, words/phrases to avoid, banned jargon
- Reference "gold standard" copy: 1–3 examples the brand considers perfectly on-voice
- Channels/markets represented: Which platforms or regions the samples come from
- Audit purpose: Pre-publish QA / new-writer onboarding / cross-channel consistency review / voice drift investigation
If no formal voice definition exists, infer a working voice profile from the gold-standard reference copy and state the inferred attributes explicitly before auditing.
Instructions
Step 1: Establish the Voice Baseline
Convert the brand voice inputs into a scorable rubric across four dimensions. Define what "on-brand" looks like for each before touching the samples:
- Personality: The consistent character (e.g., warm, irreverent, authoritative, plainspoken). What traits should come through in every piece?
- Vocabulary & Diction: Word choices, reading level, jargon tolerance, contractions, sentence rhythm. What words are signature; what words are banned?
- Tone Flexibility: How the voice shifts by context (celebratory vs. error message vs. legal) while staying recognizable. What's the acceptable range?
- Formatting & Mechanics: Punctuation habits, capitalization, emoji use, Oxford comma, sentence-case vs. title-case, exclamation tolerance.
State the baseline in 4–6 bullets so the buyer can see the standard being applied.
Step 2: Score Each Sample on the Four Dimensions
For every copy sample, assign a 1–5 score per dimension (5 = fully on-brand, 1 = clearly off-brand) and an overall score. For each dimension, cite the specific line or word driving the score — never score without evidence.
Scoring anchors:
- 5 — Indistinguishable from gold-standard copy
- 4 — On-brand with a minor, cosmetic miss
- 3 — Recognizably the brand but noticeably inconsistent
- 2 — Reads like a different brand in places
- 1 — Off-brand throughout on this dimension
Step 3: Flag Off-Brand Lines with Rewrites
Pull every line scoring the equivalent of 3 or below and present it as a fix. For each flag:
- Original: The exact off-brand line
- Issue: Which dimension it violates and why (one sentence, specific)
- On-brand rewrite: A corrected version that matches the baseline
- Rule invoked: The voice rule it now satisfies
Prioritize flags: lead with personality and vocabulary violations (they signal a different brand); mechanics/formatting fixes come last.
Step 4: Diagnose the Drift Pattern
Look across all samples for systemic causes, not one-off errors:
- Channel drift: Does one channel consistently miss (e.g., social is on-voice but product UI is stiff)?
- Market drift: Does one region's copy read as a different brand?
- Author drift: Do certain samples cluster around the same mistakes (a likely single author or agency)?
- Register creep: Is the brand slowly becoming more formal, more hype-y, or more generic over time? Name the single biggest driver of inconsistency and what to fix at the source.
Step 5: Build the Rules-of-Thumb Cheat Sheet
Distill the audit into a one-page reference any writer can apply instantly:
- 5–8 Always rules (signature moves that make it sound like the brand)
- 5–8 Never rules (specific words, structures, or habits to avoid)
- 3–4 When in doubt heuristics for tone decisions
Output Format
- Structure: Voice baseline → per-sample scorecard table → flagged lines with rewrites → drift diagnosis → cheat sheet
- Length: 900–1,300 words
- Format: Scorecard as a table (sample × four dimensions + overall); flags as structured blocks; cheat sheet as Always/Never/When-in-doubt lists
- Sections: Voice Baseline → Consistency Scorecard → Off-Brand Flags & Rewrites → Drift Diagnosis → Rules-of-Thumb Cheat Sheet
- Tone: Editorial director reviewing a team's work — direct, specific, constructive; critiques the copy, never hand-waves
Quality Checklist
Before delivering output, verify:
- [ ] Every sample has a score on all four dimensions plus an overall score
- [ ] Every score cites a specific line or word as evidence — no scores without proof
- [ ] Every flagged line includes an original, an issue, and an on-brand rewrite
- [ ] Rewrites match the stated voice baseline — verify against the gold-standard reference
- [ ] Drift diagnosis names one primary systemic cause, not just a list of individual errors
- [ ] Cheat sheet rules are specific to this brand — no generic "be consistent" advice
Notes
- When no formal style guide exists, the gold-standard reference copy is the most reliable baseline — infer from it rather than guessing at attributes.
- Voice drift is usually structural, not random: it clusters by channel, author, or market. The scorecard makes the cluster visible.
- The cheat sheet is often the highest-value output — it prevents future drift, whereas the flags only fix the current batch.
- Tone flexibility is not a license to break voice: a brand's error message and its launch announcement should still be unmistakably the same brand.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills