1---2name: repurpose3description: >-4 Turns one finished B2B SaaS article into a set of distribution assets: short LinkedIn posts built on the article's5 strongest pull quotes and statistics, an email, short video scripts and, for a framework, a carousel. Each asset is6 placed on a channel the company actually uses, aimed at one role in the buying group and one buying moment from7 company-profile.md, and spread over weeks instead of one launch day. Checks lengths, voice and that every statistic8 keeps its source. Use it after an article is signed off or published. Not for writing the article (use9 article-write), a full founder post with its own angle (use founder-post) or a month of topics (use10 content-calendar).11license: MIT12compatibility: >-13 Works in Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md. Reads the published article if the agent14 can browse; otherwise works from pasted text. Python 3 is optional, for extract candidates and length checks.15metadata:16 author: SHAKEUP17 version: "1.0.0"18 stage: make19 plugin: content20 homepage: https://www.shakeuphq.com/skills/repurpose21---22 23# Repurpose24 25Gets one article in front of the buyers it was written for, where they already spend time, for weeks after it goes live.26 27## Purpose28 29- One job: a set of assets and a placement plan from one article.30- Stage: make. It ends in its own check and a fix loop of at most 2 rounds.31- It never adds a claim the article doesn't make, never drops a statistic's source, and never places an asset on a channel the profile doesn't use.32 33## Use when / not34 35| Situation | Use |36| --- | --- |37| An article was signed off or published | this skill |38| An old article that still holds up deserves another push | this skill, with the schedule's resurface step |39| The article isn't written yet | `article-write` |40| One passage should become a full founder post with its own angle | `founder-post` |41| A landing page, pricing page or product page | not this skill: there is no editorial insight to extract. Use `ad-variants` or `creative-brief` |42| A month of topics from a chapter or theme | `content-calendar` |43 44## Inputs45 46| Input | Required | Ways to provide, in order | If missing |47| --- | --- | --- | --- |48| The article | yes | an `article-write` handoff and its file; the published URL if browsing; pasted text | ask once; `BLOCKED` without it |49| The channels the company uses | no | `## Channels` in the profile; the reader's answer | ask once; assets written with placement `not confirmed` |50| Where buyers spend time | no | the reader's evidence (sales notes, a survey, referral sources); `### Buying group` | placement from `## Channels` only, said so |51| Voice | no | `## Voice` and `### Founder voice` | plain and direct; coverage `partial` |52 53## Profile read54 55From `company-profile.md`: `## Company`, `## Buyers` and `### Buying group`, `## Buying moments`, `## Voice` (with `### Founder voice` and `### Words to avoid`), `## Claims and proof` with `### Never say`, `## Channels`, and every `## Learnings` entry for `repurpose`.56 57No profile: ask in one message for the channels the company posts on, who the buyer is, and the voice in 3 to 5 words. Say that `company-profile` saves this next time.58 59## Steps60 611. **Apply learnings.** Channel results (`fact`) re-weight the asset mix; an accepted rule ("no video until we have a presenter") removes a format.62 632. **Confirm the source.** An editorial article, guide or blog post continues. A landing, pricing, product or home page is `BLOCKED`, with the reason. An article that hasn't been signed off continues, with an open item: its claims haven't had their final check.64 653. **Read the whole article.** The strongest ideas are often not in the first section.66 674. **Find the extracts.** Run:68 69 ```70 python3 scripts/repurpose_check.py --article article.md71 ```72 73 Without Python, read every sentence and keep those of 8 to 30 words that stand alone (no opening "it" or "this" that needs the sentence before, no "as mentioned"), favouring figures with a source, contrasts and named things. Say which way you ran it. Then apply the scroll test in `references/formats.md` to each candidate and adapt as little as possible. Each statistic keeps its source (publisher and link).74 755. **Choose the asset mix.** Defaults, for active channels only:76 77 | Asset | How many | From |78 | --- | --- | --- |79 | Short post | 1 per passing extract, up to 5 | each extract |80 | Email | 1 | the strongest extract and the article's angle |81 | Video script | 2 (30 or 60 seconds) | the 2 strongest extracts that can be said aloud |82 | Carousel | 1, only if the article is a framework or a checklist | the framework |83 84 Fewer extracts pass: fewer assets. Never pad the set.85 866. **Place each asset** (`references/formats.md`, placement): one channel, one buying-group role, one buying moment, a week from the default schedule. Founder posts only when `### Founder voice` exists or the founder is a channel in `## Channels`.87 887. **Write each asset** to its format's rules: hooks from the extract, one development, one close; email subject and preview; scripts written as speech with the source said aloud. The development adds no claim, figure or personal experience the article or `### Founder voice` doesn't hold.89 908. **Check.** Run:91 92 ```93 python3 scripts/repurpose_check.py --assets assets.md94 ```95 96 Without Python, check each asset's limits by counting (150 words for a post; subject 60 and preview 40 to 90 characters; email body 250 words and 1 link; 75 or 150 spoken words for video; 10 slides of 25 words), no URL in a post body and no dashes. Then run `voice-check` (company voice for company assets, founder voice for founder posts) and `claim-check` on every asset with a figure, or apply both inline: words to avoid and point of view; every figure matches the article and keeps its named source.97 989. **Fix and recheck** (Loop), then hand over the assets and the schedule.99 100## Check and verdict101 102| Verdict | Condition (first match wins) |103| --- | --- |104| `BLOCKED` | No article, or the source is a landing, pricing, product or home page |105| `READY WITH OPEN ITEMS` | Assets pass every check, but something is open: placement `not confirmed`, the article not yet signed off, an extract whose source couldn't be confirmed (left out), a warning left after 2 rounds |106| `READY` | Every asset passes the length, voice and claim checks, each placed on an active channel with a role and a moment |107 108Named stops: `BLOCKED` asks for an article. No other stops.109 110## Loop111 1121. Collect each failed check per asset.1132. Fix only what failed: cut words, move the link to the first comment, restore a dropped source. A statistic whose source can't be confirmed is removed from the set, not reworded.1143. Rerun step 8 on the changed assets. At most 2 rounds; then hand over with anything still open, labelled.115 116## Learn117 118Append to `## Learnings` when:119 120- results come back per asset (impressions, clicks to the article, replies, video completion) against that channel's usual (a `fact`, with the source);121- the reader drops a channel or a format from the set (a `proposal` with their reason);122- an extract type keeps winning on one channel, such as statistics on the company page (a `fact` after 3 or more results).123 124Example: `- 2026-10-24 | repurpose | fact | Statistic posts drew 3x the article clicks of pull-quote posts on the company page | Why: page analytics, 9 posts, Jul to Sep 2026 | Status: noted`125 126## Output127 128```129Verdict: <READY | READY WITH OPEN ITEMS | BLOCKED> · Coverage: <complete | partial (no channels; no founder voice)> · Checks: <script | by counting>130Source: <article title, URL or file> · Extracts found: <n> (<n> pull quotes, <n> statistics)131 132| # | Extract | Type | Source |133 134| Asset | Channel | Role | Moment | Week | Check |135 136## <asset type>: <title>137<the asset; for posts, the first comment after it>138(repeat per asset)139 140Left out: <extracts or assets dropped, and why>141Open: <each open item>142Learnings applied: <list, or none>143```144 145Then the handoff block.146 147## Hand off148 149```handoff150skill: repurpose151version: 1.0.0152verdict: READY153coverage: complete154for: <article title>155outputs:156 assets: 8157 channels: [<channel>, <channel>]158 first_publish_week: <YYYY-MM-DD>159 founder_posts: [<asset title>, <asset title>]160next: [content-calendar, founder-to-ads]161open: []162```163 164`content-calendar` slots the assets into the posting calendar; `founder-to-ads` reads the founder posts once their results are in.165 166## Failure modes167 168| Problem | Do this |169| --- | --- |170| The article has no figures and few quotable lines | Fewer assets; say the article is thin for distribution and name the section that could carry one with a real example |171| A statistic in the article cites a source that no longer supports it | Leave it out of every asset and flag it for the article's owner |172| The reader wants the same post on every channel | Change the wording per channel, or keep it to 1; identical posts across channels read as automated |173| The article is gated | Assets point to the gate page; the email goes to people who haven't downloaded it |174| A channel is marked `testing` | Use it, and say the asset is part of the test |175 176## Influenced by177 178- Ross Simmonds' Create Once, Distribute Forever: distribution as the most underused part of content, remixing what already exists channel by channel, and keeping it working long after launch: https://scribemedia.com/published-books/create-once-distribute-forever179- SHAKEUP's repurposing practice: extracts that pass a scroll test without context, adapted as little as possible, one short post per extract with the link in the first comment.