Quick answer: install the free sergebulaev/linkedin-skills plugin (hooks, posts, carousels, repurposing, voice matching), add blader/humanizer to strip AI patterns, and add coreyhaines31/marketingskills if you write other marketing copy too. Paid bundles are general marketing libraries, not LinkedIn packs. Never let a tool post or engage on your account automatically: LinkedIn’s User Agreement bans most automation.
Most “AI LinkedIn post” advice skips the two parts that matter. First, which skill does the drafting well. Second, what LinkedIn actually allows once a tool starts touching your account.
This guide covers both. You get the skills worth installing (free on GitHub and paid), how each one installs, how to build hooks, carousels and document posts with them, how to turn one long article into a week of posts, how to match your own voice, which schedulers ship MCP servers, and what LinkedIn’s own documentation says about its feed and about automation.
Every star count, date and price below was checked on the live page on 2026-11-01. Where something could not be confirmed, it says so.
What are Claude Code skills for LinkedIn posts?
Direct answer: A Claude Code skill for LinkedIn posts is a folder with a SKILL.md file that tells Claude how to write, check or plan LinkedIn content. Claude loads it when your request matches its description. Skills live in ~/.claude/skills (all projects) or .claude/skills (one project), or arrive inside a plugin installed with npx skills add owner/repo or a /plugin marketplace.
That’s the whole mechanism. A skill is instructions plus optional reference files and scripts. It does not give Claude access to your LinkedIn account. Posting needs a separate connection, usually an MCP server from a scheduling tool, and we’ll get to the rules on that.
If skills are new to you, start with what Claude Code skills are and how they load and then the install guide.
Why a skill beats a saved prompt for LinkedIn
A saved prompt has to be pasted every time. A skill fires on its own when you say “write a LinkedIn post”, and it can carry things a prompt cannot hold neatly:
- A hook library with examples per format
- Your past posts as a voice reference file
- A checklist that runs before you see the draft (length, line breaks, banned phrases)
- A script that formats bold or italic text with Unicode, since LinkedIn does not render Markdown
A Moodboard newsletter quoting a Reddit user put it simply: if you need Claude to format your LinkedIn posts a certain way every time, make a skill for that (source).
Takeaway: the skill handles the writing system. Your account connection is a separate decision with separate rules.
The best Claude Code skills for LinkedIn posts in 2026
Direct answer: For most people, the free sergebulaev/linkedin-skills plugin is the best single install for LinkedIn. Add blader/humanizer to strip AI patterns, and coreyhaines31/marketingskills if you also write other marketing copy. Paid bundles make sense only if you want a wide library beyond LinkedIn.
Here’s the list. Star counts and last-commit dates come from each GitHub repo page on 2026-11-01.
1. sergebulaev/linkedin-skills (free, MIT)
- Source: github.com/sergebulaev/linkedin-skills
- Stars / last commit: about 5.5k / 2026-10-27
- What it does: 12 skills in one plugin. Post Writer, Comment Drafter, Reply Handler, Post Audit, Humanizer, Hook Extractor, Content Planner (7-day plan), Engagement Monitor, Profile Optimizer, Employee Advocacy, Repurposer and Interviewer. The Interviewer builds a “Story Bank” of your roles, numbers and opinions that the other skills read.
- Approval flow: the README says every skill shows a draft first and waits for your OK. Nothing posts without approval.
- Optional extras: Apify for reading post data and Publora for publishing. Both are third parties. Read the automation section below before you connect either.
- Install (from the README):
/plugin marketplace add sergebulaev/linkedin-skills
/plugin install linkedin-skills@linkedin-skillsOr, with the skills CLI: npx skills add sergebulaev/linkedin-skills
2. kostja94/marketing-skills, linkedin-posts skill (free)
- Source: github.com/kostja94/marketing-skills (listing on claudemarketplaces.com)
- Stars / last commit: about 1.1k / 2026-10-26
- What it does: copy for organic formats: short posts, articles, documents, polls, newsletters and reshares. It also covers making a LinkedIn presence discoverable to search and AI citation tools.
- Install (from the marketplace listing):
npx -y skills add kostja94/marketing-skills --skill linkedin-posts --agent claude-code3. github/awesome-copilot, linkedin-post-formatter (free)
- Source: github.com/github/awesome-copilot
- Stars / last commit: about 40.7k (whole repo) / 2026-10-30
- What it does: turns raw text into LinkedIn-ready formatting with Unicode bold and italics, since LinkedIn does not support Markdown.
- Install:
npx skills add github/awesome-copilot --skill linkedin-post-formatter(the repo is a large multi-agent collection; the skills CLI pulls one skill by name).
4. guia-matthieu/clawfu-skills, linkedin-post (free)
- Source: claudemarketplaces.com listing, repo github.com/guia-matthieu/clawfu-skills
- Status: GitHub marks guia-matthieu/clawfu-skills as archived (read-only; last push 1 Oct 2026). It still installs, but expect no fixes.
- Stars / last commit: about 150 / 2026-10-01
- What it does: thought-leadership posts with a GEO/SEO lean. The listing itself says it is not ideal for personal or emotional posts.
- Install (from the listing):
npx -y skills add guia-matthieu/clawfu-skills --skill linkedin-post --agent claude-code
5. coreyhaines31/marketingskills, social skill (free)
- Source: github.com/coreyhaines31/marketingskills
- Stars / last commit: about 57.8k / 2026-10-30
- What it does: the
socialskill covers creating, scheduling and improving content for LinkedIn, X, Instagram and more. The same pack hascopywriting,copy-editing,page-croandcold-email(which also touches LinkedIn outreach). - Install (from the README):
npx skills add coreyhaines31/marketingskills, or/plugin marketplace add coreyhaines31/marketingskillsthen/plugin install marketing-skills
6. Anthropic marketing plugin with brand-voice (free)
- Source: github.com/anthropics/knowledge-work-plugins
- Stars / last commit: about 34.3k / 2026-10-28
- What it does:
/draft-contentdrafts social posts among other formats./brand-reviewchecks copy against your voice and style guide. Thebrand-voiceskill documents tone and terminology. Built mainly for Cowork, and the README says it also works in Claude Code. - Install (from the README):
claude plugin marketplace add anthropics/knowledge-work-plugins, thenclaude plugin install marketing@knowledge-work-plugins
7. blader/humanizer (free)
- Source: github.com/blader/humanizer
- Stars / last commit: about 63.3k / 2026-10-28
- What it does: removes common signs of AI writing (inflated language, rule-of-three padding, em dash habits). Useful as the last pass on any LinkedIn draft.
- Install (from the README):
/plugin marketplace add blader/humanizerthen/plugin install humanizer@humanizer. On Claude Code older than 2.1.142:npx skills add blader/humanizer --global --agent claude-code
More on this category in skills for humanizing AI writing.
8. alirezarezvani/claude-skills, marketing-skills plugin (free)
- Source: github.com/alirezarezvani/claude-skills
- Stars / last commit: about 29.1k / 2026-10-31
- What it does: the README lists a 43-skill marketing plugin, including social content work. It is a broad library, not a LinkedIn specialist.
- Install (from the README):
/plugin marketplace add alirezarezvani/claude-skillsthen/plugin install marketing-skills@claude-code-skills
9. HubSpot LinkedIn post generator skill (free, email-gated)
- Source: offers.hubspot.com/linkedin-post-generator-skill
- Price: free, behind a form. On 2026-11-01 the captured page showed “Form not available”, so access may be intermittent.
- What it does: hooks, point of view, and turning messy drafts into structured posts with a CTA. HubSpot says it is built from frameworks it pulled from 50 viral posts (HubSpot’s claim, not independently checked).
- Install: download the zip, then upload in Claude.ai under Customize → Skills. For Claude Code, unzip the folder into
~/.claude/skills/.
Directories worth browsing
ComposioHQ/awesome-claude-skills (about 78.5k stars, last commit 2026-07-24) is a curated index rather than a skill. claudemarketplaces.com lists more LinkedIn skills, including linkedin-writer from ognjengt/founder-skills and linkedin-post-optimizer from onewave-ai. Directories such as skillsmp, claudskills, kissmyskills and cultofclaude index the same GitHub sources; check the repo before trusting any listing’s counts.
Free vs paid: which LinkedIn skills are worth money?
Direct answer: For LinkedIn alone, free wins. The best LinkedIn-specific skills found in this research are open source. Paid bundles are general marketing libraries where LinkedIn is one section. Buy one if you want a large library across email, SEO and ads in one place.
The paid options
- Inflectual “Master Marketing Skills Bundle” (Gumroad), $167 one-time. Contains 167 skills, including a social media group covering LinkedIn alongside email, SEO and paid ads. Price confirmed in the page’s
product:price:amountmeta on 2026-11-01. Link - SuperSkills, $50. A general skills pack. LinkedIn-specific contents were not confirmed for this post.
- Contra “Ultimate Claude SEO & AEO Skills Pack”, $39 (shown against $79). SEO-focused. Relevant only if your LinkedIn plan feeds a blog strategy; it is not a LinkedIn writing pack.
Skipped on purpose: one free “viral hooks” pack page that ran a “10 days only, $500 value” banner. That is the kind of scarcity framing this blog does not promote.
Comparison table
| Skill or pack | Free or paid | Price | LinkedIn focus | Voice matching | Install |
|---|---|---|---|---|---|
| sergebulaev/linkedin-skills | Free (MIT) | $0 | High (12 LinkedIn skills) | Yes, Story Bank + Humanizer | /plugin marketplace add |
| kostja94 linkedin-posts | Free | $0 | High (all organic formats) | Partial | npx skills add |
| awesome-copilot linkedin-post-formatter | Free | $0 | Formatting only | No | npx skills add |
| clawfu linkedin-post | Free | $0 | Thought leadership | Partial | npx skills add |
| coreyhaines31 social | Free | $0 | Medium (multi-network) | Via copy skills | npx skills add |
| Anthropic marketing + brand-voice | Free | $0 | Medium | Yes, brand-voice | claude plugin install |
| blader/humanizer | Free | $0 | Editing pass | Indirect | /plugin install |
| HubSpot Post Generator | Free (email gate) | $0 | High | Partial | Zip upload |
| Inflectual bundle | Paid | $167 one-time | One section of 167 | Not confirmed | Download |
| SuperSkills | Paid | $50 | Not confirmed | Not confirmed | Download |
| Contra SEO & AEO pack | Paid | $39 (vs $79) | Low (SEO) | No | Download |
For the wider marketing picture, see Claude Code skills for marketers.
How to write LinkedIn hooks with a Claude Code skill
Direct answer: The hook is the text above the “…see more” cut. A good skill writes several hook options per post, each in a different pattern, then lets you pick. The sergebulaev Post Writer uses a library of hook formulas, and its Hook Extractor turns any post you admire into a blank template.
Why does the hook carry so much weight? LinkedIn’s own engineering team explained it in 2020. In “Understanding dwell time to improve LinkedIn feed ranking”, the team measured dwell “on the feed” from the moment at least half of an update is visible, and dwell “after the click” on the content itself. Updates viewed for less than a threshold were treated as skips, and a model predicting the chance of a skip was used to lower ranking scores (LinkedIn Engineering, May 2020). The exact threshold was not published.
So the first two lines decide whether a reader stops. That part is documented. The specific “best hook formulas” floating around are practitioner opinion.
A hook workflow that holds up
- Give the skill the post’s single point in one sentence.
- Ask for five hooks in five patterns: a specific number, a contrarian line, a short story opening, a question the reader has asked, a before/after.
- Reject any hook that promises more than the post delivers. LinkedIn says it is reducing “engagement bait” (LinkedIn News, March 2026).
- Keep the hook under the cut. Preview the post in LinkedIn’s composer, because the cut point depends on device.
Headline craft carries over directly. See Claude Code skills for headlines.
Post formats: text, carousel and document posts
Direct answer: LinkedIn has three formats a writing skill can help with. Text posts need a hook, short paragraphs and Unicode formatting. Carousels on LinkedIn are usually PDF “document” posts, so the skill writes slide-by-slide copy and you design the slides. Document posts can also be guides or checklists. Ask the skill for a format plan before any copy.
| Format | What the skill should output | What you still do | Skill that fits |
|---|---|---|---|
| Text post | Hook, body with line breaks, one CTA, Unicode bold where needed | Read aloud, post or schedule | sergebulaev Post Writer, awesome-copilot formatter |
| Carousel (PDF document) | One idea per slide, slide headlines, 8-12 slide outline, caption | Design in Canva or Figma, export PDF | kostja94 linkedin-posts (documents) |
| Document post (guide, checklist) | Section copy, title slide, summary caption | Layout and export | kostja94 linkedin-posts |
| Poll | Question, 2-4 options, follow-up post idea | Post natively | kostja94 linkedin-posts |
| Newsletter / article | Long-form draft, headline, intro | Publish in LinkedIn editor | coreyhaines31 copywriting, Anthropic /draft-content |
Text posts
A skill should handle the plain-text rules for you: no Markdown, short lines, and Unicode for bold when you need emphasis. That’s exactly what the awesome-copilot formatter does.
Carousels and document posts
Claude Code writes copy, not finished slide design. Some setups connect a design tool through MCP, but the reliable path is copy from the skill, layout in your design tool, upload as a PDF. Keep one idea per slide and make the first slide do the hook’s job.
The sergebulaev README notes a 900 to 1,300 character “sweet spot” for text posts. That is the author’s guidance, not a LinkedIn rule.
How to turn one long-form piece into a week of LinkedIn posts
Direct answer: Feed the skill one article, transcript or newsletter. Ask it to extract five to seven distinct points, assign each a format and hook, then draft each post as a standalone that makes sense without the source. The sergebulaev Repurposer and Content Planner (7-day plan) are built for this job.
Here’s a five-day plan you can ask any of the drafting skills to produce:
| Day | Angle from the source | Format | Notes |
|---|---|---|---|
| Mon | The main claim, stated plainly | Text post | Best hook goes here |
| Tue | One step-by-step section | Document post (PDF) | Slide per step |
| Wed | A mistake or counterpoint | Text post | Invites real replies |
| Thu | A worked example or data point (sourced) | Text or carousel | Cite the source |
| Fri | A question your readers ask | Poll | Feeds next week’s ideas |
Rules that stop repurposed posts reading recycled
- Each post must stand alone. No “as I said yesterday”.
- Never paste the same paragraph twice. LinkedIn’s March 2026 update says it is “reducing recycled and click-driven posts” (source).
- Vary length and opening pattern across the week.
- Link the source in a comment or the post only if it helps the reader.
For the upstream writing, skills for content writers covers the long-form side, and email marketing skills covers turning the same piece into a newsletter.
Voice matching from your past posts
Direct answer: Export or copy 10 to 20 of your own posts into a reference file inside the skill folder. Tell the skill to study sentence length, openings, vocabulary and what you never say, then write a short voice guide. Every draft then gets checked against that guide before you see it.
A Substack guide describes this route: collect your best LinkedIn posts and build a personal skill from them (ruben.substack.com). It suggests a scraping tool for the export. Don’t. LinkedIn’s help center says it does not permit third-party crawlers, bots or extensions that scrape the site (LinkedIn Help). Copy your own posts manually, or use LinkedIn’s own data export under Settings.
A voice skill folder layout
~/.claude/skills/my-linkedin-voice/
SKILL.md # when to use, rules, checklist
references/
past-posts.md # 10-20 of your posts, pasted by hand
voice-guide.md # generated, then edited by you
never-say.md # phrases you hateTwo ready options do part of this. The sergebulaev Interviewer builds a Story Bank of your real roles and numbers so drafts stop inventing details. Anthropic’s brand-voice skill handles tone and terminology for a team.
For a full guide, see Claude Code skills for brand voice and how to build your own skill.
Scheduling tools with MCP servers
Direct answer: Buffer and Typefully both document official MCP servers that support LinkedIn. Both put posts into a queue or drafts, so you review before anything goes live. Taplio appears in MCP directories, but an official Taplio MCP doc could not be confirmed for this post.
| Tool | MCP endpoint | Review step | Plan needed | Source | |
|---|---|---|---|---|---|
| Buffer | https://mcp.buffer.com/mcp (OAuth) | Yes | Posts land in your Buffer queue as drafts or scheduled posts | ”MCP access is included on every plan, including Free” | buffer.com/mcp, developer docs |
| Typefully | https://mcp.typefully.com/mcp (OAuth) | Yes | Creates, edits and schedules drafts | Not stated for Claude Code | Typefully help |
| Taplio | Listed on mcp.so as mcp.taplio.com | LinkedIn only | Unconfirmed | Pricing not public for MCP | mcp.so listing |
Setup in Claude Code
Typefully’s help page gives this command:
claude mcp add typefully --transport http --url "https://mcp.typefully.com/mcp"Then run /mcp, pick Typefully, and finish OAuth in the browser. Buffer’s docs do not list a Claude Code command. The standard Claude Code pattern for a remote server is claude mcp add --transport http buffer https://mcp.buffer.com/mcp, then /mcp to authenticate. Check Buffer’s Integrations page if that changes.
The workflow that fits LinkedIn’s rules: skill drafts, you edit, the MCP server drops it into the scheduler, you approve in the scheduler, the scheduler publishes through LinkedIn’s API.
LinkedIn’s rules on automation: what is allowed
Direct answer: LinkedIn bans unauthorized bots, scrapers and browser extensions that automate activity on its site, including automated posting, commenting and liking. Writing drafts with AI is not banned. Publishing through a scheduler that uses LinkedIn’s official API, with you approving, is the documented route.
From LinkedIn’s “Prohibited software and extensions” help page:
- LinkedIn does not permit “any third party software, including ‘crawlers’, bots, browser plug-ins, or browser extensions” that scrape or automate activity.
- The User Agreement bars using “bots or other unauthorized automated methods to access the Services”.
- Fake engagement and tools that try to manipulate the feed are also covered.
The page warns that accounts using such tools can be restricted or shut down (source).
Where the line is
| Activity | Status | Why |
|---|---|---|
| Claude drafts a post you copy and paste | Fine | No tool touches LinkedIn |
| Scheduler posts via LinkedIn’s API after your approval | Documented route | LinkedIn’s Posts API grants w_member_social to approved apps (Microsoft Learn) |
| Browser bot that logs in and clicks “Post” | Prohibited | Unauthorized automation |
| Auto-commenting or auto-liking tools | Prohibited | LinkedIn names “comment automation” as a target (LinkedIn News) |
| Engagement pods | Targeted | LinkedIn aims to “make engagement pods ineffective” |
| Scraping posts to train a voice file | Prohibited for scrapers | Copy your own posts manually instead |
LinkedIn has no official MCP server of its own as of this writing. Every LinkedIn MCP server is third-party, so check whether it uses the official API or a logged-in browser session. Avoid the second kind.
One practical note on the sergebulaev pack: its optional Apify integration reads LinkedIn post data through scraping actors. Whether that fits your risk tolerance is your call, but it sits closer to the line than the draft-only default. The skills work without it; they ask you to paste text instead.
What LinkedIn’s own documentation says about the feed
Direct answer: LinkedIn’s published material says three things. The feed predicts skips from dwell time and ranks against them (2020). It uses large language models to understand what posts are about and match them to interests (2026). It is cutting engagement bait, recycled posts and low-substance AI content.
What LinkedIn has published
- Dwell time (2020): two dwell signals, a skip model, and a ranking score lowered in proportion to the predicted chance of a skip (LinkedIn Engineering).
- March 12, 2026 update: LinkedIn says it uses generative recommenders built on LLMs that “better understand what a post is actually about”, wants more posts with “genuine insight, actionable ideas, and thoughtful perspectives”, and is reducing “repetitive, low-substance posts and engagement bait”, including “comment to agree” prompts. It added: “We want the Feed to be relevant to your interests, not a popularity contest” (LinkedIn News).
- Recent sort: per a summary of LinkedIn’s help article, choosing Recent shows posts from people you follow in reverse chronological order without personalization (secondary source, unconfirmed on the help page itself).
Reported, not confirmed by LinkedIn here
- Press reports say LinkedIn added a “Seems like AI slop” option for members in July 2026 and uses classifiers for low-quality content (Entrepreneur). Unconfirmed against a LinkedIn primary page for this post.
- Claims that “AI posts get X% less reach” circulate widely. None traced to LinkedIn data. Treat them as unconfirmed.
What this means for your skill setup
Use skills to raise substance, not volume. A skill that checks “does this post contain one specific, true thing the reader can use?” fits what LinkedIn says it rewards better than one that pumps out seven generic posts a week.
Common mistakes with LinkedIn skills
Direct answer: The mistakes cluster in four places: letting a tool post without review, inventing numbers, recycling the same idea, and connecting scrapers or browser bots to your account.
- Auto-posting without review. Keep a human approval step every time.
- Fake specifics. Skills love “I grew 300%”. If it isn’t true, cut it. The Story Bank approach exists to supply real numbers.
- One template for every post. Readers spot the pattern by Wednesday.
- Engagement-bait CTAs. “Comment YES” is named in LinkedIn’s own list of what it reduces.
- Scraping for voice data. Copy your own posts by hand.
- Skipping the humanizer pass. Run blader/humanizer or the sergebulaev Humanizer last.
- Trusting directory counts. Check stars and last-commit date on GitHub itself.
For editing workflows, see Claude Code skills for editing and skills for copywriters.
Your setup checklist
- Install one LinkedIn drafting skill (sergebulaev/linkedin-skills or kostja94 linkedin-posts).
- Install blader/humanizer.
- Build a voice folder from 10 to 20 of your own posts, copied by hand.
- Pick a scheduler with an official MCP server (Buffer or Typefully) and connect it with OAuth.
- Ask for a five-day plan from one long-form piece.
- Review every draft in the scheduler before it goes live.
- Track what you measure yourself in LinkedIn analytics, not what a vendor promises.
FAQ
Can Claude Code post to LinkedIn directly?
Not by itself. Claude Code needs an MCP server or API connection. Buffer and Typefully document MCP servers that queue LinkedIn posts for your review.
Is using AI to write LinkedIn posts against the rules?
LinkedIn’s rules target unauthorized automation, scraping and fake engagement. Its help pages reviewed here do not ban AI-drafted text you post yourself.
Does LinkedIn penalize AI-generated posts?
LinkedIn says it is reducing low-substance posts and engagement bait. No confirmed rule penalizes AI drafting as such. Low-effort content is the stated target.
What is the best free Claude skill for LinkedIn posts?
For LinkedIn specifically, sergebulaev/linkedin-skills (MIT, 12 skills, about 5.5k stars on 2026-11-01) is the most complete free option found.
Are paid LinkedIn skill packs worth it?
For LinkedIn alone, the free options cover more ground. Paid bundles like Inflectual’s $167 pack suit people who want a broad marketing library.
How do I make Claude write LinkedIn posts in my voice?
Put 10 to 20 of your own posts in a skill’s reference folder, generate a voice guide, edit it, and have the skill check drafts against it.
Can I scrape my LinkedIn posts to train a skill?
LinkedIn prohibits third-party scrapers and crawlers. Copy your posts manually or use LinkedIn’s own data export.
Can Claude make LinkedIn carousels?
It writes the slide copy and structure. Design and PDF export happen in a design tool, then you upload it as a document post.
How long should a LinkedIn post be?
LinkedIn publishes no ideal length. The sergebulaev README suggests 900 to 1,300 characters; that is the author’s guidance.
What is dwell time on LinkedIn?
LinkedIn’s engineering team defines feed dwell as time while at least half of an update is visible, and separate dwell after a click. Short dwell counts as a skip.
Is Taplio’s MCP server official?
A Taplio MCP server appears on mcp.so, but this post could not confirm an official Taplio doc. Check with Taplio before connecting it.
Is Buffer’s MCP server free?
Buffer’s MCP page says “MCP access is included on every plan, including Free.”
Can a skill reply to LinkedIn comments for me?
The sergebulaev Reply Handler drafts replies for you to post. Auto-commenting tools fall under what LinkedIn calls comment automation.
Do engagement pods still work?
LinkedIn said in March 2026 that it aims to “make engagement pods ineffective”. Effectiveness after that is unconfirmed.
Where do Claude Code skills get installed?
In ~/.claude/skills for all projects, .claude/skills for one project, or via a plugin with npx skills add or /plugin.
Do skills work in Claude.ai as well as Claude Code?
Many do. The sergebulaev README lists Claude.ai, Desktop and Code; it notes skills and plugins need a paid Claude plan with code execution.
Sources and methodology
Checked live on 2026-11-01. Star counts and last-commit dates come from each repo’s GitHub page and commit feed. Install commands come from each repo README or the marketplace listing named next to it. The Inflectual price comes from the Gumroad page’s product:price:amount meta tag. SuperSkills and Contra pack prices match the figures verified on 2026-11-01; their LinkedIn-specific contents were not confirmed.
- LinkedIn Help, Prohibited software and extensions: https://www.linkedin.com/help/linkedin/answer/a1341387
- LinkedIn Engineering, Understanding dwell time (May 12, 2020): https://www.linkedin.com/blog/engineering/feed/understanding-feed-dwell-time
- LinkedIn News, Improving the Feed (March 12, 2026): https://news.linkedin.com/2026/ImprovingTheFeed
- Microsoft Learn, LinkedIn Posts API: https://learn.microsoft.com/en-us/linkedin/marketing/integrations/community-management/shares/posts-api
- Buffer MCP: https://buffer.com/mcp and https://developers.buffer.com/guides/integrations/mcp.html
- Typefully MCP: https://support.typefully.com/en/articles/13128440-typefully-mcp-server
No skill in this post was tested by the author for this article. Descriptions come from each project’s own documentation.
Conclusion: your first week
Install one drafting skill and the humanizer today. Build your voice folder by hand tomorrow. Then take your best long-form piece and ask for five posts, one per weekday, each in a different format. Connect a scheduler with an official MCP server and approve every post yourself.
That setup stays inside LinkedIn’s published rules and aims at what LinkedIn says its feed wants: specific, useful posts that hold attention. Next, read Claude Code skills for SEO to point the same long-form work at search.
