Skip to content

borghei/Claude-Skills / tools/linkedin/linkedin-content-repurposer

linkedin-content-repurposer

Verified

Turns something made for another channel (a thread, a video or talk transcript, a blog post, a newsletter) into a post that reads as native to LinkedIn, without changing what the source says. Use when existing material should become a LinkedIn post rather than writing one from a blank page.

references/source-format-playbooks.md

View on GitHub
# Source Format Playbooks

One playbook per kind of source. Each says what the format does to a piece of
writing, what the analyser will flag, and how to rebuild it. They share one
method, described first.

## Contents

1. The shared method
2. Thread
3. Talk or video transcript
4. Article or blog post
5. Newsletter issue
6. Podcast or interview appearance
7. Short note or caption
8. Slide deck without a script
9. Choosing between one post and several
10. Quick reference

## 1. The shared method

Every channel shapes writing to its own constraints. A thread is cut into
segments because of a length cap. A talk repeats itself because listeners
cannot re-read. An article has headings because it expects to be scanned.
Those shapes are not part of the idea. Repurposing means separating the two.

1. **Find the spine.** The spine is the single point the post will carry,
   stated in one sentence the author would defend. It normally contains a
   concrete thing (a number, a date, a named case) and a turn (what was
   believed before, what was found instead).
2. **List the carry-over.** The facts that must arrive intact: figures with
   units and periods, names, the sequence of events.
3. **List the cut.** Everything the old channel required and the new one does
   not.
4. **Close the source.** Write the post from the spine and the carry-over
   list, not from the source text. This is the step that prevents paste and
   trim.
5. **Reopen the source and check.** Run the fidelity check. Compare every
   figure by eye as well.

The analyser supports steps 1 to 3. It does not write, and it cannot know
which point the author cares about most.

## 2. Thread

**What the format did.** Split the argument into numbered segments, each
written to stand alone in a feed. Put a promise in the first segment and an
ask to repost or follow in the last. Used handles and tags for reach.

**What the analyser flags.** RP-001 numbering and thread markers, RP-002
handles, RP-003 hashtag clusters, RP-005 repost and follow asks, RP-004 links,
RP-020 an opening that announces a thread.

**What survives.** Threads are usually already near post length and often
already concrete. The content is frequently fine. The seams are the problem.

**Rebuild.**

| Step | Do |
|------|----|
| Opening | The first segment sells the thread. Discard it. Open on the finding itself, usually found in the middle or near the end |
| Joins | Segments were written to work alone, so each restates context. Remove the restatements; let one paragraph lead into the next |
| Order | Threads delay the payoff to keep people scrolling. A post can state the outcome early and then explain it |
| Mentions | Replace handles with names where the person should be credited; drop the rest |
| Ending | Replace the repost ask with one question or one suggestion the reader can act on |
| Link | Decide deliberately; see the native-fit guide |

**Common mistake.** Deleting the numbers and leaving nine two-sentence
paragraphs, each still beginning as though the reader had just arrived.

## 3. Talk or video transcript

**What the format did.** Speech is redundant by design: points are announced,
made, and repeated. The speaker refers to things the audience can see. Filler
words and false starts are normal. An automatic transcript adds timestamps,
speaker labels and mis-heard words.

**What the analyser flags.** RP-006 timestamps and labels, RP-007 spoken
filler, RP-008 references to slides, the room or the audience, RP-024 long
sentences, RP-020 an opening made of greetings.

**What survives.** Talks are rich in material: speakers tell stories and give
numbers they would not bother to write down. But the transcript sentences
themselves almost never survive. Treat a transcript as notes, not as a draft.

**Rebuild.**

| Step | Do |
|------|----|
| Check the words | Automatic transcripts mis-hear names and numbers. Confirm every figure with the speaker or the slides before using it |
| Pick one section | A talk has an arc of several arguments. Choose one. The spine candidates usually cluster in two or three places |
| Replace what was shown | "As you can see on the chart" must become the thing the chart showed, in words, with its numbers |
| Write, do not edit | Do not clean transcript sentences. State the idea fresh in written prose |
| Keep one spoken line | A phrase the speaker actually said, kept verbatim, carries their voice. One is enough |
| Other speakers | Remove the host's and questioners' words unless credited with consent |

**Common mistake.** Running a transcript through a tidy-up and publishing the
result: it reads like someone talking at you without pausing.

**A caution on recorded material.** The transcript must come from the user as
text. If only a recording exists, the user produces the transcript by whatever
means they choose and supplies the file.

## 4. Article or blog post

**What the format did.** Organised several points under headings, introduced
them, cross-referenced them, concluded them. Used tables, images, code blocks
and footnotes. Assumed a reader who chose to open it and has minutes to spend.

**What the analyser flags.** RP-009 headings, cross-references and "in this
article", RP-010 tables, images and code, RP-023 several separable points,
RP-025 long paragraphs.

**What survives.** Each section is typically a post. The introduction and
conclusion are typically nothing.

**Rebuild.**

| Step | Do |
|------|----|
| Split | Use the units list. One section, one post |
| Self-sufficiency | Remove every "as mentioned above". If a section depends on an earlier one, restate the dependency in a clause or choose a different section |
| Tables | Turn the two most telling cells into a sentence. A full table belongs in an attached document, if anywhere |
| Opening | The article's title and introduction describe the article. Open on the section's finding |
| Depth | Articles explain; posts point. Keep the claim, the evidence and the turn. Leave the full method for the article |
| Link back | Optional and deliberate. The post must make sense to someone who never clicks |

**Common mistake.** A post that lists the article's headings as bullets and
ends by directing the reader to the article. That is an advertisement for the
article, not a post.

## 5. Newsletter issue

**What the format did.** Addressed subscribers who already know the author.
Mixed several items: an essay, links, updates, a sign-off. Used a familiar,
ongoing tone ("as I said last week").

**What the analyser flags.** RP-009 reader address and cross-references,
RP-011 references to earlier issues, RP-005 subscribe asks, RP-004 links.

**Rebuild.**

| Step | Do |
|------|----|
| Choose one item | Usually the essay or one argument from it. Link round-ups do not convert |
| Remove assumed familiarity | A subscriber knows your product and last week's issue. A feed reader does not. Add the one line of context they need |
| Tone | Newsletters can ramble pleasantly. Tighten, but keep the voice |
| Subscription ask | If it stays at all, it is the single ask of the post and goes at the end. Do not combine it with a question |

**Common mistake.** Posting the issue's opening paragraph and "read the rest
in my newsletter". The reader has been given nothing.

## 6. Podcast or interview appearance

**What the format did.** A conversation: the best material comes out as
answers to someone else's questions, interleaved with the host's own views.

**Special risk.** Attribution. Once labels are stripped, the host's sentences
look like the guest's.

**Rebuild.**

| Step | Do |
|------|----|
| Mark speakers first | Before any cutting, confirm which passages are the author's own |
| Use answers, not exchanges | Take one answer and rebuild it as a first-person account. Do not reproduce the back-and-forth |
| Credit the prompt if it mattered | If the host's question is what made the answer interesting, say so and name them, with consent |
| Do not quote yourself | "As I said on the show" is channel history. State the point |

## 7. Short note or caption

**What the format did.** Compressed a thought into a line or two, often for a
visual it accompanied.

**What the analyser says.** Move: expand. Often RP-022 if there is no figure
or date at all.

**Rebuild.** Expansion is the one case where repurposing needs new material.
Padding a short note with general remarks produces exactly the generic post
this suite exists to prevent. Get from the author: when this happened, the
case behind it, a number if one exists, and what they thought before. If a
story bank is available, look there first. If nothing concrete exists, the
note is not yet a post.

## 8. Slide deck without a script

Slides alone are headings without sentences. Do not infer the argument from
bullet points: the reasoning lived in what the speaker said. Ask the author to
talk through the three slides that matter, or to supply notes, and treat that
as the source. A slide's numbers still need their units, periods and basis.

## 9. Choosing between one post and several

| Sign | Reading |
|------|---------|
| Source inside the band, one argument | One post |
| Source above the band, one argument with supporting detail | One post, compressed |
| Source well above the band, sections that each contain a figure and a turn | Several posts, one per section |
| Sections that only make sense in sequence | One post on the conclusion; the sequence stays in the source |
| Several posts possible | Space them out. Publishing a talk as five consecutive posts turns the feed into a serial nobody subscribed to |

Each post in a set must pass the stranger test: does it make complete sense
to someone who sees only this one and never sees the others?

## 10. Quick reference

| Source | Biggest artefact | First move | Typical yield |
|--------|------------------|------------|---------------|
| Thread | Segment seams and the repost ask | New opening from the middle | 1 post |
| Talk transcript | Spoken redundancy and slide references | Pick one section | 2 to 4 posts |
| Article | Headings and cross-references | Split by section | 1 per substantial section |
| Newsletter | Assumed familiarity | Choose one item | 1 post |
| Interview | Mixed speakers | Mark whose words are whose | 1 to 3 posts |
| Short note | Nothing to cut, little to carry | Get detail from the author | 0 or 1 |
| Slides only | No sentences | Get the script | 0 until the author talks it through |

Yields are rough expectations from the structure of each format, not
measurements.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall