Marca personal
Repurpose content
propost-repurpose-content
Cuándo ayuda
Turns one long source the person provides (article, blog post, talk transcript, podcast notes, newsletter, webinar, or text they paste from a URL) into several posts across their networks in ProPost, with network-specific cuts, using create_draft with per_channel and copy_post_to_channel, and never attributing anything to the person that the source doesn't say. Use for "repurpose this", "turn my article into LinkedIn posts", "make posts from my podcast", "cut this talk into posts", "get more out of my newsletter", "post about this blog post".
Los skills funcionan con el conector de ProPost, disponible en Pro y Premium. Configurar el conector
Takes one piece of the person's own long-form material and cuts it into a small set of distinct posts, each fitted to its network, each faithful to the source. Quality here is selection and fidelity, not volume.
Before you start
get_account: drafts left,autonomy, time zone.list_channels: networks,max_length, image/video requirements,needs_reconnect.list_personas: voice per channel.list_posts(published, limit 20): so the new posts do not repeat what they just said.- Get the source text in the conversation. The assistant has no tool to fetch a URL through ProPost. If they only give a link, ask them to paste the text (or use your own browsing if you have it, then treat what you read as the source). Confirm they wrote it or have the right to quote it.
Workflow
- Read the whole source first. Extract, in a working list: the main claim, 5 to 10 distinct points, every number, every named example, 3 to 5 quotable sentences (verbatim), any story or scene, any step-by-step.
- Choose 3 to 7 post angles (not more than the source supports). Each angle has one point, from one part of the source. Pick for variety: claim, story, how-to steps, number, myth, quote, question. See
references/cuts.md. - Propose a table before drafting: post, angle, source section it draws on (quote the first few words), format, channels, uses AI draft or not, suggested timing. Say the AI draft count.
- After a yes, write.
You write the text yourself from the source:create_draftwithchannel_ids,contentfor the main cut andper_channelfor the shorter networks. No AI draft used. Best for fidelity.
Orgenerate_draftwith anideathat carries the source passage and the point (uses 1 AI draft per distinct voice). Then read it against the source and fix anything added or exaggerated withupdate_post.
To reuse a finished post on another network:copy_post_to_channel(it condenses to the limit automatically,shortened: true); then check the shortened text still says what the original said. - Fidelity pass on every post: each claim and number must appear in the source. Fix or delete what does not. Quotes must be verbatim.
- Check:
check_post; fix. - Spread over time: don't post all from one source in a day. Use
suggest_posting_timesandadd_to_queue(schedule level) over 1 to 3 weeks, orschedule_postwith exact times. Drafts level: leave as drafts and give the proposed dates. - Report: list of posts with the source passage each came from, what needs their confirmation.
Network-specific cuts
- LinkedIn: the story or argument in 900 to 1,500 characters, one idea, link to the source in the first comment or last line if they want traffic (the person adds it; link text goes in the post only if they asked).
- X, Threads, Bluesky, Mastodon: the single sharpest sentence or number; for a thread, 3 to 5 short parts, each standing on its own. Bluesky and X are tight on length: check
max_length. - Instagram, Facebook: a short caption that adds one new line to an image or video; for carousels, a slide-by-slide breakdown the person turns into images (
attach_imagewithadd: truefor multi-image posts). - Google Business, Telegram: practical and local, one update, one next step.
Writing guidance
- Change the entry, not the content: each post opens from a different part of the source. Don't summarise the whole thing five times.
- Lead with the strongest sentence from the source, usually not its introduction.
- Cut transition and throat-clearing. Posts start in the middle of the argument.
- One source reference is enough ("From my talk at X", "From last week's newsletter") and only if true.
- Keep their words where the source is good; edit for length, not for style. Don't turn a plain sentence into a slogan.
Example. Source (podcast notes): "We lost our largest customer in month 14 because we hadn't spoken to them in seven months." Post:
We lost our biggest customer in month 14. Not because of price or a competitor. We hadn't spoken to them in seven months. Now every account over [amount] gets a call every 6 weeks, whether or not anything is wrong.
(The "[amount]" marker stays until the person supplies it; the rule itself must be in the source.)
Guardrails
- Never claim the person said or did anything the source does not say. No new statistics, no invented quotes, no "I" statements about events not in the source, no inferred opinions.
- Don't blend in outside facts. If you add context from elsewhere, label it and ask first.
- Quotes from guests or third parties: only if the source includes them and the person confirms they may be shared; credit them.
- Material the person did not write or lacks rights to: ask first.
- No private data about others from transcripts (names, personal details) without consent.
- Don't
publish_now; don't schedule posts the person has not seen.
If something fails
quota_exceeded: write all posts yourself withcreate_draft.too_long: edit theper_channeltext down;check_post.empty_content: the text was blank; resend with content.weekly_limit: keep the remaining posts as drafts; list them.channel_needs_reconnect: skip that channel; say which.- Anything else: see
../CONVENTIONS.md.
Archivos de referencia
Cutting one source into postsreferences/cuts.md
A quick menu of angles. Pick only those the source genuinely supports.
| Cut | What to take from the source | Typical network |
|---|---|---|
| The claim | The main argument in 3 to 6 short paragraphs | |
| The number | One figure from the source with what it means | X, LinkedIn |
| The story | A scene or anecdote told start to finish | LinkedIn, Instagram caption |
| The steps | A how-to from the source, 3 to 5 steps | LinkedIn, carousel |
| The quote | One verbatim sentence, with a line of context | X, Threads, Bluesky |
| The myth | A belief the source disputes, and its replacement | LinkedIn, X |
| The question | An open question the source raises | Threads, Bluesky, LinkedIn |
| The teaser | One surprising point that sends readers to the full piece | Any |
Rules of thumb:
- One source gives 3 to 7 good posts; a 1,500 word article rarely supports more than 4.
- Each post answers a different question. If two do, merge them.
- Keep a source-line table while working: post, passage it draws from. If a sentence has no passage, it does not go in.
Cómo añadir un skill
- Claude (web y escritorio): descarga el skill en .zip y añádelo en Ajustes, Capacidades, Skills.
- Claude Code: descomprímelo en ~/.claude/skills/ (o en .claude/skills/ de tu proyecto).
- Otros asistentes: abre el SKILL.md y pégalo en las instrucciones personalizadas o el conocimiento del proyecto.
Los skills están escritos en inglés. Tu asistente te responderá y escribirá los posts en tu idioma.