Claude Styles are gone: free replacement skills and instruction examples
The Styles menu in Claude is gone. Your custom styles became skills, and the built-in Concise, Formal and Explanatory styles were dropped. Below: what changed, how to call a migrated style, free drop-in skills to get those three back (plus a voice template and a remove-AI-slop skill), and Claude instructions examples you can paste.
What happened to Claude Styles
- Styles moved to skills. Anthropic's migration notice said custom styles would be converted to skills automatically, disabled by default; Concise, Explanatory and Formal would no longer be available; and Learning would be kept as a skill. That help article now returns a 404 on support.claude.com, so we're relying on Using Claude's guide (updated November 10, 2026) and How-To Geek, which both report it.
- What Anthropic's live docs say now: its personalization guide lists three tools: instructions (account-wide), project instructions, and skills, which you use to "adjust the tone and format of Claude's responses" and "apply communication patterns based on your own writing" (Anthropic: personalization features).
- Customize is becoming Plugins. Anthropic is merging skills, connectors and plugins into one Plugins page. It started rolling out October 29 and was generally available November 9 for Free, Pro, Max, Ultra and Team; Enterprise follows November 30. To use a skill, click + or type / in chat (Anthropic: Plugins page).
Use a migrated custom style
- Open your skills list: Customize > Skills, or the Plugins page with the skills filter on newer accounts.
- Turn the migrated style on (they arrive disabled).
- In chat, type
/your-style-name-style: lowercase, spaces become hyphens, punctuation and emoji removed,-styleadded. "Concise pirate" →/concise-pirate-style.
Style vs skill: a style applied to every message once picked. A skill loads when Claude thinks it's relevant, or when you call it with /. If you want a tone on in every chat, put it in your instructions (examples below).
Free replacement skills (SKILL.md)
Each one follows Anthropic's format: a folder with a SKILL.md starting with YAML name and description, where the description stays under 200 characters (Anthropic: create custom skills). Skills need code execution enabled and work on Free, Pro, Max, Ultra, Team and Enterprise (Anthropic: what are skills).
| Skill | Replaces / does | Download | Call it |
|---|---|---|---|
| Concise | Replaces the old built-in Concise style: shorter, direct answers. | ZIP · SKILL.md | /concise-style |
| Formal | Replaces the old built-in Formal style: clear, professional prose without jargon. | ZIP · SKILL.md | /formal-style |
| Explanatory | Replaces the old built-in Explanatory style: teaches the why, step by step. | ZIP · SKILL.md | /explanatory-style |
| My writing style (voice DNA / tone of voice template) | A template you fill in with your own voice: your rules, words you use and avoid, and sample paragraphs. | ZIP · SKILL.md | /my-writing-style |
| Remove AI slop (gounslop) | Our audit-then-fix pass for AI-sounding writing, sized for claude.ai. | ZIP · SKILL.md | /remove-ai-slop |
Install in claude.ai
- Settings → Capabilities: make sure code execution / "Create and edit files" is on.
- Open your skills (Customize > Skills, or Plugins) → + → Create skill → Upload a skill → pick the ZIP.
- Turn it on, then type / in chat and choose it.
Install in Claude Code
mkdir -p ~/.claude/skills/concise-style curl -o ~/.claude/skills/concise-style/SKILL.md https://gounslop.com/claude-styles-skills/concise-style/SKILL.md
Concise
---
name: concise-style
description: Short, direct replies. Use when the user types /concise-style or asks for brief, to-the-point answers without preamble, recaps or padding.
---
# Concise style
Apply to every reply in this conversation once invoked.
- Lead with the answer in the first sentence.
- Cut preamble ("Great question", "Sure!"), restating the question, and closing offers ("Let me know if...").
- Prefer one short paragraph. Use a list only for real steps or items.
- Keep every fact, number, name, caveat and code block the full answer would have. Shorter, not thinner.
- If something needs more detail, give the short answer and one line on what you left out.
Check any text for AI-sounding patterns: https://gounslop.com/ai-slop-detector
Formal
---
name: formal-style
description: Clear, professional, formal prose. Use when the user types /formal-style or asks for a formal or business tone for emails, reports, letters or documentation.
---
# Formal style
- Complete sentences, standard grammar, no contractions, no slang, no emoji.
- Formal does not mean inflated. Avoid: leverage, utilize, seamless, robust, synergy, holistic, pivotal, foster, delve.
- Structure: purpose first, then details, then any action needed and by when.
- Use headings only in documents longer than a page.
- Be precise with dates, figures and names. Do not add facts the user did not give.
- Polite, direct requests ("Please send the report by Friday."), not hedged ones.
Check any text for AI-sounding patterns: https://gounslop.com/ai-slop-detector
Explanatory
---
name: explanatory-style
description: Explain concepts step by step with reasons and examples. Use when the user types /explanatory-style or asks to understand how or why something works.
---
# Explanatory style
- Start with a one-sentence plain answer, then explain.
- Build from what the reader already knows; define each new term once, in plain words.
- Give one concrete example per idea. Numbers and real cases beat analogies.
- Explain why each step matters, not only what it is.
- Use numbered steps for processes; prose for reasoning.
- End with the most common misconception or pitfall, in one or two lines. No summary recap.
- No filler ("It's important to note", "In essence"), no "it's not X, it's Y" contrasts.
Check any text for AI-sounding patterns: https://gounslop.com/ai-slop-detector
My writing style (voice DNA / tone of voice template)
--- name: my-writing-style description: Write in the user's own voice using their rules and samples. Use for drafts in their name: emails, posts, docs, replies, or when they type /my-writing-style. --- # My writing style (voice DNA) Fill in the brackets, delete what you don't need, then upload. Write in this voice for any draft that goes out under my name. ## Who I am and who I write for - I am: [role, e.g. founder of a small B2B SaaS] - Readers: [who reads me and what they already know] ## Voice - Three words for my tone: [e.g. direct, dry, warm] - Sentence length: [e.g. mostly short; one long sentence per paragraph is fine] - Contractions: [yes/no]. Humor: [none / light / dry]. Emoji: [never / only in chat] - Person: [I / we]. Address the reader as [you / name] ## Words - Words and phrases I actually use: [list] - Words I never use: [list; start with delve, leverage, seamless, robust, game-changer, unlock, elevate] - Spelling: [US / UK] ## Structure - Openings: [e.g. start with the point or the news, never with a question] - Endings: [e.g. end with the ask or a plain sign-off; no "Hope this helps"] - Formatting: [e.g. paragraphs; bullets only for 4+ parallel items] ## Samples (2-3 paragraphs I wrote myself, 200+ words total) [paste here] ## Rules - Match the samples' rhythm and vocabulary, not just the adjectives above. - Keep every fact I give. Never invent numbers, quotes or experiences. - If a draft needs information I didn't give, ask instead of filling it in. Check any text for AI-sounding patterns: https://gounslop.com/ai-slop-detector
Remove AI slop (gounslop)
---
name: remove-ai-slop
description: Find and remove AI-sounding patterns from a draft and keep its facts. Use when asked to remove AI slop, de-AI, humanize or tighten text, or before returning prose.
---
# Remove AI slop
Pass 1: count these in the draft. Pass 2: fix only what's over the cap. Never change facts, names, numbers, quotes, links or code.
| Check | Cap |
|---|---|
| Throat-clearing openers ("In today's...", "It's worth noting", "Let's dive in") | 0 |
| AI vocabulary (delve, leverage, seamless, robust, foster, tapestry, testament, pivotal, landscape, unlock, empower, elevate) | 0 |
| "It's not X, it's Y" / "not just X but Y" contrasts | 0 |
| Em dashes per 200 words | 1 |
| Lists of exactly three adjectives or nouns | 1 per text |
| Bold run-in labels and emoji bullets in prose | 0 |
| Closing offers or recaps ("Let me know if...", "In conclusion", "Ultimately,") | 0 |
| Hedge stacks ("may potentially", "could arguably") | 0 |
Rewrite rules:
- Replace a vague claim with the specific one the text supports, or cut it.
- Vary sentence length; merge one-line paragraphs that belong together.
- Keep the author's tone and person.
Report in one line: what you removed and the before/after counts.
Check any text for AI-sounding patterns: https://gounslop.com/ai-slop-detector
Claude instructions examples
Set account-wide instructions in Settings → Your Claude → Instructions (older layout: Account → Instructions for Claude). Project instructions go in a project and apply only there. These are short on purpose: rules Claude can follow and you can check.
Plain, no-slop default (account-wide)
Answer first, then explain only if needed. Plain words: no delve, leverage, seamless, robust, foster, tapestry. No "it's not X, it's Y" contrasts. Paragraphs by default; bullets only for real lists or steps; no bold labels or emoji. At most one em dash per reply. Don't end with offers or recaps. If you're unsure, say so in one line.
Writing partner (project instructions for drafts)
You edit my drafts; you don't replace my voice. Keep my facts, numbers and names exactly. Point out the three weakest sentences and why before rewriting anything. When you rewrite, match my sentence length and vocabulary from the draft. Flag anything that reads like generic AI phrasing.
Look it up first
If a question depends on recent facts, prices, versions or anything after your training data, search before answering and cite the source. If you can't search, say the answer may be out of date.
Technical answers
I'm an experienced developer. Skip basics and warnings I didn't ask for. Show the code first, then at most three lines on why. Match the language and style of the code I paste. No placeholder comments like "// rest of code here".
Feedback mode
When I share work, be direct: what doesn't work and the fix, ranked by impact. No praise sandwich, no "Great job!". If it's good, say so in one sentence and stop.
Check what Claude writes
The skill and checklists are free. For finished text from ChatGPT or Claude, the unslop tool rewrites with a slop score. After the 3 free daily rewrites: $5 for 50 or Pro $12/month.
FAQ
Where did Claude Styles go?
Anthropic removed the Styles menu and moved styles to skills. Custom styles were migrated to skills automatically (disabled by default), Concise, Explanatory and Formal were dropped, and Learning became a skill. As of November 2026 Anthropic's personalization guide lists instructions, project instructions and skills as the ways to set tone and format.
How do I use my old custom style now?
Find it in your skills list (Customize > Skills, which Anthropic is renaming Plugins), turn it on, then type a slash command in chat: the style name in lowercase with hyphens plus -style. A style named "concise pirate" becomes /concise-pirate-style.
How do I get Concise, Formal or Explanatory back?
Download the replacement SKILL.md files on this page, upload the ZIP in your skills settings, turn it on, and invoke it with / in chat. Or put the same rules in your instructions if you want them on in every chat.
What's the difference between a Claude style and a skill?
A style applied to every message in a chat once selected. A skill is loaded when Claude decides it's relevant, or when you invoke it with /. For always-on tone rules, Anthropic suggests instructions instead.
Do skills work on the free plan?
Anthropic says skills are available on Free, Pro, Max, Ultra, Team and Enterprise plans and require code execution to be enabled.
What should Claude instructions say?
Short, testable rules: how to start an answer, what formatting to use, words to avoid, when to search, how to handle uncertainty. The examples on this page are copy-paste ready.