算法艺术创作
algorithmic-art
通过代码和可调参数创作生成艺术与交互式视觉效果。
linkedin-post【设计与创意】Write a high-performing LinkedIn post in the user's brand voice using Hook-Story-Offer (Russell Brunson) and the Curiosity Gap hook method (George Loewenstein). Every post starts by selecting a named hook formula - specific number + surprising result, contrarian statement, personal confession, direct challenge, or before/after. Use when the user asks…
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Writes LinkedIn posts that get read, saved, and reshared. Applies the Hook-Story-Offer structure (Russell Brunson) with the Curiosity Gap hook method (George Loewenstein). Every post creates a specific information gap in the hook, closes it in the story, and ends with one clear action.
Brunson's sequence, used here as three stages with a job, a failure mode and a test each. Write them in this order and do not start the story before the hook is settled.
| Stage | Its one job | It has failed when | The test to run |
|---|---|---|---|
| Hook | Buy the click on "see more" | The reader can scroll past without wondering anything | Cover everything but line 1. Is there still a question in the reader's head? |
| Story | Earn trust by closing the gap the hook opened | It restates the hook at greater length, or answers a different question | Read the hook, then the last line of the story. Did the promise get paid? |
| Offer | Convert attention into one action | It ends with "thoughts?" and nothing else | Underline the imperative verb. If there is no verb, there is no offer |
HOOK: stop the scroll. Create a pattern interrupt. The hook must create enough curiosity or pain recognition that the reader clicks "see more." The hook fails if someone can scroll past without wondering what comes next.
STORY: make the hook real. Provide context, the journey, the insight. This is where trust is built. Show, don't tell. Specific details, numbers, named situations.
OFFER: what should the reader DO? The offer can be: click a link, follow, save the post, reply with a word, or believe something new. One action only. Low friction. "What do you think?" is not an offer - it is an afterthought.
Note: "offer" in LinkedIn context is NOT always a sales pitch. The offer is whatever action you want the reader to take - including "bookmark this framework" or "reconsider this belief."
Match the offer to the goal. Pick from this table rather than defaulting to a link.
| Goal | Offer to use | Offer to avoid | Reason |
|---|---|---|---|
| Build authority | "Save this" or "steal this checklist" | An external link | A save is a low-friction commitment that signals value |
| Generate leads | One link, named plainly | A vague "DM me" | The reader should know exactly what is on the other side |
| Start a conversation | A specific question with a constrained answer | "What do you think?" | An open question is work, a constrained one is a reply |
| Recruit | "Reply with the word [X]" | "Apply here" | Lower the first step, qualify in the DM |
| Announce | One link plus one sentence on who it is for | Three links | Split attention halves both |
Loewenstein's information-gap account of curiosity: curiosity is the feeling of a gap between what you know and what you want to know, and it is strongest when you know just enough to see exactly what is missing. A hook that tells you nothing creates no gap, and a hook that tells you everything closes it before you scroll.
Best LinkedIn hooks create a specific information gap between what the reader knows and what the post promises to reveal. The gap must be:
Score the hook before writing anything else. Three axes, 0 to 2 each.
| Axis | 0 | 1 | 2 |
|---|---|---|---|
| Specificity | Names a topic | Names a category of answer | Names one thing, countable or singular |
| Closeability | The answer is guessable from the hook | Partly guessable | Only the post can answer it |
| Relevance | Interesting to anyone | Interesting to the industry | Interesting to this named persona this week |
Threshold: ship at 5 or 6. At 4, rewrite the hook. At 3 or below, the problem is the angle, not the wording, so go back to step 3 and pick a different formula. Never rewrite a 3 into a 5 by adding adjectives, that produces clickbait, which is a gap the story cannot pay off.
Hook quality test: does the hook create a specific gap the reader needs to close? If someone can guess the answer from the hook alone, it is not a gap - it is a headline.
Pick one before writing. Name it in the process.
Load context. Read knowledge/brand/voice.md. If missing, stop and tell the user to run /brand-context.
Read 3-5 high-performing past posts from knowledge/content-library/ if they exist. Mirror cadence and rhythm, not topic.
Select a hook formula. Pick one of the 5 formulas. Name it explicitly before writing. Do not skip this step.
Hook formula selected: [formula name]
Gap this hook creates: [one sentence - what does the reader not know yet?]
Gap score: specificity _/2 closeability _/2 relevance _/2 = _/6
If the score is below 5, do not proceed. Rewrite at 4, change formula at 3 or below.
Write the hook. First line decides everything. Rules:
Write the story. 3-7 short paragraphs using Hook-Story-Offer structure:
knowledge/brand/voice.md says they are part of the voice.Write the offer. One line. One action. Low friction.
Self-check before showing. Every item is checkable by counting or quoting the draft.
Framework:
Sourcing:
knowledge/content-library/case-studies/ or in the
user's brief. List each number and its source[NEEDS INPUT: <what>] rather than a placeholder figureFormat:
Save to output/linkedin-post/<DD-MM-YYYY>-<slug>.md with frontmatter:
---
format: linkedin-post
topic: <topic>
author: <founder|cmo|company>
goal: <goal>
hook-formula: <formula name>
gap-score: <n>/6
words: <count>
created: DD-MM-YYYY
---
Offer 2 hook variants unless the user only wanted one. The hook is the highest-leverage edit point - a different hook formula on the same content can double or halve performance.
| Situation | Best formula |
|---|---|
| You have a specific metric result | Formula 1: Specific number + surprising result |
| You want to challenge a common belief | Formula 2: Contrarian statement |
| You changed your own mind about something | Formula 3: Personal confession |
| You see a widespread mistake | Formula 4: Direct challenge |
| You have a transformation story | Formula 5: Before/after without the how |
knowledge/services/
and knowledge/content-library/case-studies/ only, both of which step 1 loads.knowledge/content-library/case-studies/ and confirm it
exists there. If it does not, use hook formula 2, 3 or 4 - never 1 or 5. Those two formulas are
built around a specific number, so choosing them without one forces invention.Do not publish, and say which condition fired.
knowledge/ or the brief. Never invent
one. Switch to hook formula 2, 3 or 4, or emit [NEEDS INPUT: the actual number] and stop.[APPROVAL NEEDED: <name>], do not present it as ready.Ranked worst first. The first three block publication.
knowledge/brand/voice.md says so./thought-leadership-writer when the idea needs 1200 words and a defended thesis, not a post/content-repurposer to turn one article, webinar or case study into a run of posts/social-calendar for the posting plan and pillar mix this post should slot into/content-writer for the same idea in a different format, such as an email or blog intro/copy-review to grade a post the user already drafted rather than writing a new one/case-study-writer when the post keeps needing a result that has not been documented yet/ab-copy-writer when two hooks are both credible and the choice should be tested/brand-context first, whenever knowledge/brand/voice.md does not exist yet/kpi-review to check afterwards whether this format is actually working for the accountThese are real limitations of this skill. Ask the user or emit [NEEDS INPUT: <what>] rather
than filling them in.
knowledge/learnings.md or the account's analytics disagree with this file, the
account data wins.knowledge/.Any cadence, character count, best-day, format spec or benchmark in this skill is a starting default recorded at authoring time, not a published platform rule. Two obligations:
knowledge/learnings.md or the
team's analytics contradict a default here, follow the data and say so in the output.Never present a default from this file to a client as though the platform published it. That is the claim their in-house specialist corrects in the meeting, and the correction discredits everything else in the document.