Claude LinkedIn Skill: Full SKILL.md + Workflow

A Claude LinkedIn skill is a folder with one file — SKILL.md — that teaches Claude your voice, your post formats, and your rules once, so every post after that sounds like you instead of a generic AI paraphrase of your notes. Paste raw thoughts, a meeting transcript, or a client win, and Claude turns it into a finished post that matches how you actually write, not how a template writes.

This guide gives you the complete SKILL.md code, explains why a LinkedIn skill beats a one-off prompt, and lays out a weekly workflow for turning one piece of source material into a week of posts. If you have not built a skill before, start with what Claude skills are and come back — the concepts below assume you know a skill is just a folder Claude loads when relevant.

Why a skill beats a prompt for LinkedIn content

A prompt only lives for one conversation. A skill lives in Claude permanently and loads itself automatically whenever you mention LinkedIn, a post, or content — so you stop re-explaining your voice, your hooks, your formatting rules, and your no-go words every single time.

The gap shows up fastest in voice drift. Ask Claude cold to "write me a LinkedIn post about this," and you get competent, forgettable copy — hedged, generic, sprinkled with "game-changer" and "unlock." Give Claude a skill that spells out your actual sentence rhythm, your real opening lines, and the three things you never say, and the tenth post still sounds like the first. That consistency is the entire point of building a skill instead of collecting prompts — it is the difference between briefing a ghostwriter once and re-briefing them from scratch every Monday.

Skills exist across the Claude app, Claude Code, and Cowork — build the LinkedIn skill once and it works wherever you write. For the broader case on what else skills can standardize, see Claude for marketers, which covers ad copy, email, and SEO workflows built the same way.

The complete SKILL.md, ready to copy

Below is a full LinkedIn content skill. Copy it into a file named SKILL.md, put that file inside a folder called linkedin-posts, and fill in the bracketed sections with your own voice details. The name in the frontmatter must match the folder name exactly, and skill names cannot contain the words "claude" or "anthropic" — those are reserved.

---
name: linkedin-posts
description: Writes LinkedIn posts in my voice from raw notes, transcripts, or ideas. Use whenever the task involves drafting, editing, or repurposing content for LinkedIn.
---

# LinkedIn Post Writer

## My voice
- Tone: [e.g. "direct, a little dry, no corporate cheerfulness"]
- Sentence rhythm: short sentences. One idea per sentence. Break
  up anything over 20 words.
- First person only. Never "we" unless describing my team by name.
- Words I never use: game-changer, unlock, delve, leverage, synergy,
  disrupt, revolutionize, "in today's fast-paced world"
- Words/phrases that ARE my voice: [3-5 phrases you actually say]

## Post structure
1. Hook — the first line must work with feed truncation. No
   questions as openers ("Ever wonder..."). Lead with a claim,
   a number, or a scene.
2. Body — 3-6 short paragraphs, one line each is fine. Tell the
   story or make the argument in the order I'd say it out loud,
   not in "problem/solution/CTA" template order.
3. Turn — one line that reframes or contradicts the obvious take.
   This is the line people screenshot.
4. Close — a takeaway or a genuine question, not a canned CTA.
   No "Thoughts?" No "Agree?"

## Formatting rules
- 150-250 words for a standard post, up to 400 for a story post.
- Line breaks between every 1-2 sentences (LinkedIn readability).
- Maximum 3 hashtags, placed at the very end, only if they are
  specific (not #motivation #leadership #growth).
- No emojis unless I explicitly ask for them.

## When given source material
If I paste a transcript, article, or notes:
1. Extract the 2-3 most specific, least obvious claims — skip
   anything that sounds like every other post in this niche.
2. Pick ONE claim per post. Do not try to cover everything.
3. Ask me which claim to run with if there are multiple strong
   options, rather than guessing.

## Output format
Give me:
1. The post, ready to paste
2. Two alternative hook lines
3. One line on why you chose this angle over the others

The frontmatter is doing real work here: the description field is what Claude scans at the start of a session to decide whether this skill applies, so it needs to name both what the skill does and when to reach for it. Everything below the frontmatter is the instruction body Claude only reads once it decides the skill is relevant — that is the progressive-disclosure design that makes skills cheap to keep installed even when you are not using them.

Making the voice section actually work

Most LinkedIn skills fail at the voice section because people describe their voice the way a résumé describes a personality — "authentic, insightful, engaging." That tells Claude nothing. What works is pasting 3-5 of your best past posts directly into the skill (or a linked reference file) and asking Claude to derive the actual patterns before you write the voice rules yourself.

Here are 5 of my best-performing LinkedIn posts. Analyze them and
write me a "voice" section I can drop into a SKILL.md file.

Cover: typical sentence length, how I open posts, how I close them,
words/phrases I repeat, what I never do (no hashtag spam, no
corporate language, etc.), and structural patterns across all five.

[paste 5 posts]

Run that once, drop the output into the voice section of the skill above, and every post after it starts from your actual patterns instead of a guess. This is the same principle behind Claude design skills — specificity in the skill file is what separates output that looks AI-written from output that looks like you wrote it on a good day.

The weekly content workflow

A skill only pays off if you feed it consistently. Here is the weekly loop that turns one source into a full week of posts, using the skill above every time.

The repurposing pipeline: one asset, five formats

The same skill setup handles repurposing, which is where most of the time savings actually live. A single long-form asset — a blog post, a webinar recording, a client case study — can feed a week of LinkedIn content without ever feeling copy-pasted, as long as you tell Claude to reshape it for the platform rather than trim it.

Using my linkedin-posts skill, turn this source material into 3
distinct LinkedIn posts. Do not summarize the same way three times —
each post should take a different angle:

POST 1: The single most counterintuitive finding, told as a
mini-story with a specific before/after.
POST 2: A mistake or wrong assumption the source corrects — written
as "I used to think X, then Y happened."
POST 3: A practical takeaway framed as a short list, but written in
my voice, not a generic listicle format.

SOURCE MATERIAL:
"""
[paste the blog post, transcript, or case study here]
"""

This is the same repurposing logic covered in more depth in Claude for marketers — the skill just makes it repeatable without re-explaining your voice in every prompt.

Common mistakes that make skill-written posts sound fake

Three mistakes account for most LinkedIn posts that read as obviously AI-written, and all three are fixable in the skill file itself rather than by editing every post after the fact.

• No banned-words list.Without an explicit list, Claude drifts toward safe, generic phrasing like "game-changer" and "unlock." Naming the words you never say does more than any tone instruction.
• Vague voice descriptions."Professional but approachable" describes ten thousand LinkedIn accounts. Pasting real past posts and deriving actual patterns, as described above, is what fixes this.
• Covering too many claims per post.A post trying to make five points makes none of them memorably. The skill above forces one claim per post — resist the urge to override that when you have "so much to say."

Where to go from here

Build the skill above this week, run it against one real piece of source material, and edit in the 10% only you can add before you post. That single loop — one skill, one weekly source, one honest edit pass — is what turns LinkedIn from a chore into a five-minute habit. Once the LinkedIn skill is working, the same pattern extends to design work and anywhere else your voice or standards need to stay consistent — see our roundup of the best Claude skills for what to build next, and what are Claude skills if you want the fundamentals first.

Frequently Asked Questions

Do I need Claude Pro to use a LinkedIn skill?

Skills work across the Claude app, Claude Code, and Cowork. Check claude.com for which plans include skill uploads on your surface of choice, as availability by plan can change.

Can I use one skill for LinkedIn and other platforms?

You can, but it fights the point of a skill. LinkedIn and X reward different structures and lengths. Build a separate, focused skill per platform rather than one skill with branching instructions for every channel — a small, specific skill outperforms one giant skill trying to do everything.

Will posts written with a skill sound like AI wrote them?

Only if the skill file is vague. The fixes are specific: ban the generic words, derive your voice from actual past posts instead of adjectives, and always do a final edit pass to add the detail only you know. A well-built skill removes the blank-page problem, not the editing step.

How is a skill different from just saving a good prompt?

A saved prompt is text you paste manually every time. A skill is a folder Claude loads automatically once it decides the task matches — it also supports linked reference files and examples the skill can pull in only when needed, which a single prompt cannot do.