| name | linkedin |
| description | Write a LinkedIn post in the user's voice. Use when the user asks to draft a LinkedIn post, hook, or long-form social caption. Pulls voice, stories, beliefs, stats, humour, and vocabulary from the me/ context files and follows the structure of examples/linkedin-example.md. |
/linkedin — LinkedIn post in your voice
Use this skill any time the user asks you to draft a LinkedIn post, hook, carousel caption, or long-form social post. Output is ready to paste into LinkedIn.
Step 1 — Load the user's voice
Always read these files first (they live in the me/ folder at the root of this project). Do not guess the user's voice — pull from source:
me/identity.md — who they are, what they sellme/tone.md — how they writeme/vocabulary.md — words they use + words to avoidme/humour.md — how they jokeme/beliefs.md — their hot takes / positionsme/stories.md — anecdotes they tellme/data.md — numbers they name-dropme/analogies.md — metaphors they use
Also read examples/linkedin-example.md for the canonical post structure (if it exists).
Step 2 — Ask only what you can't infer
If the user gave you a topic, run with it. Only ask if it's genuinely ambiguous (e.g., "post about AI" — clarify angle). Default to action over questioning.
Step 3 — Post structure
- Bold unicode headline (𝐁𝐎𝐋𝐃) — the big claim or result. Specifics, dollar amounts, percentages.
- Hook line — one punchy line that tightens the headline. Use a 👉 arrow if that matches the user's style.
- Pattern interrupt — short line that creates tension or curiosity.
- Story bridge — transition into the story (e.g., "Let me explain how I got there:").
- Narrative body — short paragraphs, 1–3 sentences each. Heavy use of line breaks. One thought per line. Use a real story from
me/stories.md when the topic allows.
- Payoff / lesson — the one-line takeaway. Often maps to a belief from
me/beliefs.md.
- Numbered list of specifics — 3–5 items with real numbers pulled from
me/data.md. Never invent stats.
- Soft close / CTA — what the reader does next. No hard pitch unless requested.
Step 4 — Voice rules (non-negotiable)
- Match the user's default voice from
me/tone.md. Never sound generic.
- Use their sentence rhythm, not yours.
- Drop in their real numbers from
me/data.md — never invent stats.
- Self-deprecation, humour, and profanity should match the levels defined in
me/humour.md and me/vocabulary.md.
- Banned words from
me/vocabulary.md are a hard filter. Scan the draft and remove every instance before delivering.
- End with a clear next step or a quietly confident line. No "what do you think?" beg-for-engagement closers unless the user asks for one.
Step 5 — Banned words (hard filter)
Before returning the draft, scan against me/vocabulary.md § "Banned words" section and remove every instance. Common AI-sounding offenders to watch for:
delve, tapestry, realm, testament, navigate (as verb), robust, leverage (as verb), synergy, holistic, ecosystem, bespoke, elevate, unlock, streamline, game-changer, moving forward, circle back, touch base, low-hanging fruit.
Also avoid emoji spam unless the user's style explicitly uses them. Match their emoji usage from me/tone.md.
Step 6 — Length & format
- Default: ~150–300 words. Long enough to tell a story, short enough to read on a phone.
- Short hook post: ~60–120 words — bold headline + hook + 2–3 paragraphs + CTA.
- Long-form / announcement: 300–500 words — full story arc.
Default to long-form when the topic is a launch, announcement, or result reveal. Default to short when it's an opinion or quick take.
Step 7 — Deliver
Output the post as a plain markdown code block (so the user can copy it cleanly). Below the post, in 1–2 lines, note:
- Which story / stat you anchored on
- One alternate hook the user can swap in
Do not write a preamble. Do not explain the post. Just deliver it.
Topic-to-story mapping
What success looks like
Reader can tell within the first line who wrote this. The post lands a real number, tells one true story, and ends with a clear next step. Nothing in it sounds like it was written by an AI.
