# Linkedin Profile > Audits and rewrites the nine sections of a person's public LinkedIn page from material they can prove, headline and About first. Triggers: optimize my LinkedIn profile page, rewrite my headline, my About section reads flat. Not for building the editorial profile file (use linkedin-setup), not for writing a post (use linkedin-post). - Skill: `alexis-morain/linkedin-profile` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add alexis-morain/linkedin-profile` - Raw SKILL.md: https://api.skillmd.com/api/skills/alexis-morain/linkedin-profile/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: alexis-morain (https://skillmd.com/u/alexis-morain) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/alexis-morain/linkedin-profile --- # The public page > **Paths in this file are relative to the bundle root**, the directory holding > the router `SKILL.md`, not to this skill's own directory. The bundle installs > as one unit so that `references/`, `locales/` and `lib/` resolve the same way > from every skill in it. Every post this engine writes sends readers to one place: the person's public LinkedIn page. A reader who liked a post lands there and decides in a few seconds whether the author is somebody or noise. This skill makes the page carry the same argument as the posts, under the same constraint: **it cannot claim anything the person cannot prove.** Two different things share the word "profile" and this skill touches only one of them. The editorial profile is the Profile section of the material, which feeds every skill; `linkedin-setup` owns it. The **public page** is what LinkedIn shows the world; this skill owns that. ## Before anything 1. Read the material's `## Status` block. If `filled: no`, stop and offer `linkedin-setup`. A page rewritten without a source of proven facts is copywriting, and this bundle does not do copywriting. 2. Read its Profile, Pillars and Voice sections. 3. Read `locales//style.md` and `interview.md`. Fall back to `locales/en` and say so. 4. Ask the person to paste their current page, section by section, or export it (LinkedIn offers a PDF of the full profile). **The engine never scrapes and never touches the platform.** The person pastes, the person applies. 5. If a page from a previous run was kept, read it and diff: what was adopted, what drifted since. ## The audit Run the promise in reverse. Every claim currently on the page gets one of three verdicts: - **Traces**: backed by the material, by a pasted post, or by something the person says right now. Keep, maybe sharpen. - **Provable but absent from the material**: true, but the material does not carry it. It goes into the material first, with its source. The page never holds facts the profile does not. - **Traces to nothing**: the inherited superlative, the borrowed metric, the "passionate about" filler. Flag it, quoted, and propose removal. A page is the one place a stale exaggeration sits in public for years. Announce the verdicts in one compact list before touching anything. ## The interview Three intents, `references/interview-intents.md`, set C. Wording in `locales//interview.md`. One question at a time, and skip anything the material already answers, which is most things. This interview is short on purpose: the material is supposed to exist already. ## The nine sections Worked in this order. The first three carry the argument; the rest follow from them. | # | Section | Its one job | |---|---|---| | 1 | Headline | Travels with the name into every feed, comment and search result. Who it serves plus one provable claim, in the buyer's words. Not a stack of job titles, not a string of pipes and keywords. | | 2 | About | The three lines above the fold decide whether anyone expands. Open on the thesis, then the two or three facts the `proof-pick` intent selected, then who this is for, then the next step from the `reader-next-step` intent. First person, in the output language, in the voice the Voice section describes. | | 3 | Featured | Two or three items, each one a proof, not a decoration. Built from the `proof-pick` selection; the best measured post in the ledger belongs here. | | 4 | Experience | The current role written as outcomes with sourced numbers, consistent with "what I sell". Past roles compressed to what explains the present one. | | 5 | Skills | The terms from the `buyer-words` intent, the ones clients actually type and say. A skills list is a search surface, not a trophy shelf. | | 6 | Recommendations | Requested from the public, named references from the material only. "Names I must never cite" applies here exactly as it does in a post. Draft the ask in the output language if the person wants one. | | 7 | Photo | A recent face, framed close enough to read in a feed avatar. Mechanism only; this skill has no opinion on style. | | 8 | Banner | The one static surface for a stated promise or a number the person can prove. Text content is this skill's job; design is not. | | 9 | Activity | Not written here at all. The page inherits it from the posting loop, and a strong page above a dead feed reads as abandoned. If cadence is the weak point, say so and route to `linkedin-post`. | ## Writing Open the generation with an explicit output-language directive, one line, before anything else. Note that headline and About may be in a different language from the interview; the `output_language_default` axis applies to the page too. Then, per section worked: 1. **Headline: three proposals, three different leads.** Each anchored to a fact, with the fact named underneath. Not three phrasings of one idea. 2. **About: one draft**, structured as above. **Every factual sentence prints its source beside it**, in a list under the draft: the sentence, then the line of the material or of this conversation it came from, quoted. A sentence with nothing under it is one nobody can check, and it comes out of the draft rather than into the page. Run the deterministic pass on it, it is prose: ```bash python3 lib/lint.py --lang - < about.txt ``` 3. **The rest as concrete edits**: current text, proposed text, one line of why. Small diffs get applied; rewrites get postponed. **No keyword stuffing, anywhere.** A headline written for an algorithm reads like one, and the reader it costs is worth more than the search impression it buys. The `buyer-words` intent puts real search terms in naturally or they stay out. ## Archive Emit it the way every session in this bundle ends, as one block of the lines that changed, for the person to paste and keep: ``` PAGE UPDATE updated: 2026-09-09 Headline: About: ``` Nothing is written here. A tier that holds files may store it, and the next run starts by diffing against whatever the person brings back. One mechanism across the three skills, so somebody who has closed one session already knows how this one ends. The person applies the changes on LinkedIn by hand. When they say it is done, update `updated`. If they adopted a variant of a proposal, archive what they actually applied, not what was proposed. ## Hard rules - **No invented number, no borrowed client, no inherited superlative.** Every claim traces to the material or to this conversation, **and the trace is printed, not asserted**: name the source next to the claim so the person can check it. What traces only to this conversation goes back into the material with its source. - **Names that are not public stay off the page.** Same list as everywhere else. - **The page never contradicts the posts.** If the headline promises what the pillars never deliver, the fix is the headline, not a new pillar. - **Do not fill a section on the person's behalf and move on.** Proposals, then their call. The page carries their name in a way a post does not: a post scrolls away, the page stays. - **Style**: `locales//style.md` applies to the About in full.