# Linkedin Post Writer > Writes professional short-form posts of 120 to 300 words carrying one idea and one piece of evidence in the author's own voice, with a first line that is a plain statement rather than a hook, real paragraphs rather than the broken-line format, and a close that states the point once instead of asking for comments. Also adapts a newsletter issue, a paper, a launch or a piece of work into a post that stands alone for readers who will never click. Use this skill whenever someone asks to write a LinkedIn post, announce something publicly, share an article, a paper, a milestone or a new role, "post about this", "put something up about it", turn an issue or a report into a post, or asks whether a draft post is any good; and for any short professional social writing under a person's own name. - Skill: `ingridleiria/linkedin-post-writer` (Agent Skill) - Install (CLI): `npx skillmds@latest add ingridleiria/linkedin-post-writer` - Raw SKILL.md: https://api.skillmd.com/api/skills/ingridleiria/linkedin-post-writer/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: ingridleiria (https://skillmd.com/u/ingridleiria) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/ingridleiria/linkedin-post-writer --- # LinkedIn Post Writer The feed has taught people a format, and the format is now the problem. A five-word opening line. A blank line. A story with a reversal. Three bullets of lesson. A question at the end asking whether you agree. It was optimised for a ranking system, it spread because it worked for a while, and readers have since learned to recognise it in under a second and scroll past. The cost is not lost reach, which is what authors worry about. It is that the audience worth having is small and knows the field. Perhaps forty people in a professional network can actually judge the author's work and might hire them, fund them, cite them, or send an introduction. Those readers use a post as evidence about how the author thinks. A post shaped for the algorithm tells them the author was thinking about the algorithm, and it takes several good posts to undo the impression of one that farmed for comments. The alternative is not modesty. It is a post that states one specific thing, supports it with one piece of evidence, and stops. That reads as a person talking, and it gets read to the end by the people who matter. ## When to use this, and when not to Use it to write a post from scratch, to adapt a newsletter issue or a paper or an internal analysis into one, to announce something the author has made, to comment on something public, and to review a draft the author is unsure about. Do not use it for long form. An idea that needs evidence, a counter-argument and a close belongs to `newsletter-post-writer`, and the post here is the pointer to it, not a compressed version of it. Do not use it for a message sent to a named person, which is `outreach-email` whether it goes by email or by direct message. The two are opposite: a post says one thing to many people who owe the author nothing; a message makes one request of one person. Do not use it for an organisation's voice, a product launch from a company account, or an all-staff communication, which is `ceo-communications`. Do not use it for a public statement during an incident, which is `crisis-and-incident-comms` and has its own approval path. Do not use it to write a bid, a proposal or an application dressed as a post. ## What you need before starting **The one idea, in a sentence.** A post with two ideas is two posts, and the second one always weakens the first because the reader is choosing which to remember. Missing: write both sentences down and ask the author which one they would keep if they could only publish one. **The evidence.** A number, a case from the author's own work, a document, a paper that has been verified. Without it the post is an opinion, and the feed is not short of those. Missing: either find one or do not post. This is the input whose absence most often means the post should not exist yet. **The reason to post now.** Something published, shipped, decided, observed this week, or a public event to respond to. Posts anchored in time read as real; posts with no occasion read as content production. Missing: ask what changed this week. If nothing did, the honest answer is to wait. **The author's voice.** Their previous posts and their longer writing: sentence length, whether they use the first person freely, whether they use humour, what they never do. Missing: write plainly, avoid all stylistic risk, and tell the author the register is neutral rather than theirs. **Who the audience actually is.** Not the follower count. The two or three groups whose opinion matters: peers in the field, potential clients or collaborators, current colleagues, a specific hiring or funding audience. The composition decides how much context a post needs. Missing: assume informed peers and add one clause of context for outsiders. **What may be said.** Employer policy, client confidentiality, embargo dates, coauthors who have not seen the result, anything under review. Missing: ask before drafting, not after. A post is public the moment it is published and cannot be recalled by deleting it. **Where the link goes, if there is one.** The exact destination, checked. Missing: draft without the link and mark the place; never publish a post pointing at a page that does not yet exist. ## The method 1. **Write the one idea as a sentence a person would say aloud.** If the sentence needs a subordinate clause to make sense, it is probably two ideas. Keep the harder one. 2. **Establish the occasion and state it early.** The reader should know within two sentences why this is being said today. An observation from Tuesday, a paper published this morning, a repository made public, a decision the author changed their mind about after a conversation. 3. **Pick the type and take its requirements.** Use the table below. The type decides what the post must contain, and mixing types is the commonest cause of a post that says nothing: an announcement with a lesson bolted on reads as a humble brag, and an observation with a link appended reads as an advertisement. 4. **Write the first line as a plain statement.** It is the line that shows before the fold, so it does the work of a headline, but it must be a sentence the author would say to a colleague. Not a question, not "Unpopular opinion", not a single word for effect, not a number teasing a list. The test: read the line alone and ask whether it is a claim someone could disagree with or a fact someone could learn. If it is neither, it is a hook, and hooks are what the reader is scrolling past. 5. **Write the body as two to four real paragraphs.** The evidence with its source, the reasoning in one or two moves, and what follows from it. Full sentences, and paragraphs of more than one line. The one-sentence-per-line format signals a post built for scanning rather than reading, and readers who know the field read it as a tell. 6. **Attach the evidence in the sentence that needs it.** A number with its unit and its period; a case with enough detail to be checkable; a paper with the author and year in the text. "A client" is acceptable where confidentiality requires it; "a Fortune 500 company" is decoration and reads as a way of not saying anything. 7. **Close by stating the point once.** No call to action, no "Agree?", no "Thoughts?". A question belongs only where the author genuinely wants an answer and will reply to the people who give one, and it should be specific enough that a well informed person could answer it in a sentence. 8. **Place the link by consequence, not by superstition.** If the link is the reason for the post, put it in the body and name what it is: readers who would follow it should not have to hunt. If the post stands on its own and the link is supporting material, the first comment is fine, mentioned in the post so nobody misses it. Do not build the whole post around avoiding an outbound link; a post that hides its own subject to game distribution has already lost the reader it wanted. 9. **Run the confidentiality pass before the style pass.** Reread as though you were the client, the coauthor, the employer, or the person described. Anything identifying that has not been cleared is either changed or removed. Anonymising by removing the name but keeping three identifying details does not work. 10. **Cut to length.** Under 120 words is a caption and usually means the evidence was left out. Over 300 words is a newsletter issue in the wrong place and should be published as one, with a post pointing to it. Cut in this order: the second example, the qualification nobody will dispute, the sentence explaining why the topic matters, and any sentence that restates the first line. 11. **Hand it to `human-voice-editor`, then read it aloud once.** The tells in short form are more conspicuous than in long form because there is nothing else on the page. 12. **Decide the follow-up before publishing.** Who the author will reply to, and what they will do with a good reply. A post that produces four thoughtful comments and no response from the author is worse than no post, and this is the part most often skipped. ## Post types and what each requires | Type | Must contain | Must not contain | Words | | Announcement: something made, published or launched | what it is in the first sentence, why it exists, what a reader gets from it, the link | a lesson tacked on, a list of tags, "humbled" | 130 to 220 | | Observation from the author's own work | the thing observed, the number or fact behind it, what it implies | a constructed story with a twist | 150 to 300 | | Adaptation of a longer piece | the single strongest claim, one number or quote from it, the link | a summary of the whole piece | 150 to 250 | | Reaction to something public | what the thing says in one sentence, where the author extends or disagrees, the link to it | a rewrite of the original with no addition | 120 to 250 | | New role or a change | the role and the date, what the author will be doing, one sentence of thanks to named people | a career narrative, a list of tagged accounts | 120 to 180 | The observation is the most readable type and the most underused, because it requires the author to have noticed something specific rather than to have an opinion about a trend. ## Worked example **Situation.** Hana Lindqvist runs operations at a fifty-person environmental consultancy. Over eighteen months she had changed how the firm hired analysts, moving from a CV screen to a two-hour paid work sample, and she had eighteen months of data: 34 hires under the old process and 29 under the new, with a first-year departure rate that had fallen from 8 of 34 to 2 of 29. She wanted to post about it. **Task.** One post, in her voice, aimed at the twenty or so operations and people leads in her network whose opinion she respected, plus the analysts she hoped would apply next. It had to be accurate enough that a statistician among her contacts would not object. **Action.** The first draft was 310 words in the standard shape: a one-line hook reading "We stopped reading CVs.", a blank line, a short story about a candidate who had been rejected on paper and became a strong hire, and three lessons in bullets. It was abandoned for two reasons. Two of the three lessons were generic and true of any hiring process, which meant the post had three ideas and therefore none. And the hook was not something Hana would say; the firm had not stopped reading CVs, it had stopped screening on them, and the line was false in the service of being punchy. The rewrite started from the one idea: a paid work sample changed who they hired, not just how well they hired, because it moved the decision from a proxy to the work itself. The evidence needed care. The raw comparison, 8 of 34 against 2 of 29, is a striking drop and is not by itself strong evidence, because the labour market and the firm both changed over eighteen months. The wrong turn worth recording is that the second draft led with the departure rate as though it settled the question. A colleague pointed out that the sample was small enough that the difference could easily be noise. Rather than dropping the number, the post kept it and stated the limitation in the same breath: small numbers over a changing period, offered as the reason the firm kept going rather than as proof it worked. The stronger evidence turned out to be a different fact she nearly left out, that 11 of the 29 hires under the new process had backgrounds the old CV screen would have filtered out, which is a claim about composition and does not depend on the outcome data at all. The first line became a plain statement of that composition fact. The body ran three paragraphs: what the work sample is and that it is paid, the composition number with the departure figures and their limitation, and what she would do differently, which was to shorten the sample from two hours to ninety minutes because two hours excluded people with caring responsibilities. The close stated the point once and named the thing she was still unsure about. There was no link, because there was nothing to link to, and no question at the end. Final length was 240 words, in four paragraphs. The confidentiality pass changed nothing, since no candidate was identifiable, and removed one sentence about a competitor's process that was accurate but not hers to tell. **Result.** Reach was unremarkable and Hana did not track it closely. What she could observe was that the post drew eleven comments, of which four were substantive, including one from a person she had wanted to meet for a year and one from a researcher who sent two papers on work sample validity that she had not seen. Two analysts referenced the post in applications the following month. She replied to every substantive comment within a day, which she credits more than the post itself. The honest uncertainty: whether the hiring change works remains unproven, and the post says so. She expects to know in another two years, and she has committed to posting the follow-up whichever way it goes, which is a harder promise than it sounds. ### A second scenario, where it goes differently The same author adapting a 2,400-word newsletter issue into a post faces a different problem, and the instinct that ruins it is summarising. A summary of a long piece is a list of claims with the evidence removed, which is precisely the shape of a generic post. The rule instead is to take the single strongest claim from the issue, give it the one number that supports it, and let the post stand alone for the large majority of readers who will never click through. The post is not an advertisement for the issue; it is a smaller, complete thing that happens to have a longer version. Two other things change. The link goes in the body rather than the first comment, because here the link is genuinely the point for the minority who want the argument in full. And the first line is drawn from the middle of the issue rather than its opening, because a newsletter opens on a scene that needs several paragraphs to pay off, and a post has no room for a slow start. What changed: the constraint is no longer finding evidence but discarding it. Adaptation is a subtractive job, and the failure mode is a post that gestures at four claims because the author could not bear to lose three of them. ## Output ``` POST [120 to 300 words. First line a plain statement. Two to four real paragraphs. Evidence in the sentence that needs it. Close states the point once.] FIRST COMMENT (if the link goes there) [One line naming what the link is, and the link.] NOTES FOR THE AUTHOR Type: [from the table] One idea: [the sentence] Evidence: [what it is, and how strong it actually is] Occasion: [why now] Cleared: [confidentiality checked, and anything removed] Follow-up: [who to reply to, and what to do with a good reply] ``` For a review of an existing draft, return the marked diagnosis before the rewrite: | Check | Verdict | Note | | First line: statement or hook | | | | Number of ideas | | | | Evidence present and specific | | | | Format: paragraphs or broken lines | | | | Stock vocabulary and engagement bait | | | | Close: point or request for comments | | | | Length | | | | Anything not cleared to say | | | ## Failure modes **The hook that is not true.** Recognise it when the first line overstates what the post then explains. The line is doing distribution work at the cost of accuracy, and the people whose opinion matters notice the gap between line one and paragraph two. Rewrite the line as the plain version of the claim. **Three lessons, two of them generic.** Recognise it by asking whether each lesson would survive being read by someone in a different industry. If it would, it is not a finding, it is a platitude. Keep one and cut the rest. **Evidence with no limitation.** Recognise it when a small sample or a short period is presented as settled. Stating the limitation in the same sentence costs nine words and is the difference between a post a specialist respects and one they quietly discount. **The constructed story.** Recognise it by the reversal: a candidate everyone dismissed, a meeting that changed everything, a stranger with a lesson. Readers' detectors for these are well trained. Use the thing that actually happened, including its lack of a clean ending. **Broken-line formatting.** Recognise it because more than half the lines are single sentences with blank lines between them. Reassemble into paragraphs and see whether anything is left; often the format was concealing that there were only two sentences of content. **Announcement with a lesson attached.** Recognise it when a post about something the author made pivots to what the journey taught them. Announce the thing and stop. The lesson is a separate post, later. **Tagging and hashtag inflation.** Recognise it by tagged people who are not part of the story and by more than two hashtags. Both read as distribution mechanics and both cost credibility with the readers who matter. **Posting something not cleared.** Recognise it by a detail the author knows but the public does not. Run the confidentiality pass before the style pass, every time, because a good post is harder to hold back than a bad one. ## Edge cases **The result is not the author's alone.** Coauthors, a team, a client. Name them, and get agreement before publishing rather than crediting them afterwards. A post claiming a team's work in the first person is remembered by the team. **Announcing bad news or a failure.** State what happened, what it cost, what changed as a result, and nothing else. No lesson framing, no redemption arc. Where anyone else is involved, the post covers the author's own conduct only. **The author is job hunting.** The strongest post is a piece of work, not a statement of availability. One line at the end saying what they are looking for is enough, and it is more effective after a post that demonstrated something than as a post on its own. **The topic is contested.** Decide before drafting whether the author intends to argue it in the comments. If not, do not post it, because a contested claim left unanswered reads as abandoned. If yes, state the position plainly, include the strongest counter-argument in the post, and set aside time. **The evidence is confidential in full.** Post the shape without the numbers: the direction, the mechanism, the decision it led to. A directional claim honestly labelled is publishable; a fabricated round number is not. **A long thread would be better.** If the idea genuinely needs 700 words, it is a newsletter issue or an article. Write it there and post the pointer. ## Quality bar - One idea, one piece of evidence, one point at the close. - The first line is a plain statement the author would say aloud, and it is true of what follows. - Real paragraphs, no line broken for effect. - The evidence carries its number, period and limitation, or names its source. - Nothing in the post exists to game distribution: no bait question, no false hook, no tag or hashtag that is not part of the story. - Between 120 and 300 words. - Everything in it is cleared to say, and anyone whose work it describes has agreed. - The author knows who they will reply to. ## Adapting this to your context The defaults come from one LinkedIn network, writing about operations and research for a few hundred readers who mostly know the author. - **The word bands.** 120 to 300 overall, narrower per type. They follow where LinkedIn truncates. Re-derive them from where your own posts get cut, and allow for languages that run longer than English. - **Platform mechanics.** The before-the-fold line, the link in the first comment and the two hashtag ceiling are current conventions, not principles. Check what your platform truncates and how it treats outbound links. - **The post types.** Five rows: announcement, observation, adaptation, reaction, role change. Add or drop rows for what you publish. A clinician bound by patient confidentiality needs a type this table does not have. - **The evidence standard.** The worked example argues from 34 hires against 29 and states the limitation in the same sentence. Where your evidence is qualitative, name the case, the period and how many instances you saw. - **The confidentiality pass.** Written for client and employer confidentiality. Add your own gates: a publication embargo, ethics approval, a coauthor's agreement, an employer's social policy. - **What not to change.** One idea, one piece of evidence carrying its own limitation, and a first line that is true of what follows. The rest is platform convention. ## Related skills `newsletter-post-writer` produces the long-form piece this post points to, and hands over the single strongest claim rather than a summary. `human-voice-editor` is the final pass on every draft here. `literature-verification` applies to any post citing research, before drafting. `breaktalk-brand` supplies the card format, sign-off and image treatment where the post carries that identity. `outreach-email` covers the message to one named person, which this deliberately excludes, and `ceo-communications` covers the organisational voice.