Repurpose post

Adapt a finished blog post into the requested channel formats. Every output must stand alone: the reader shouldn't need the original post to get the value. Done means each requested format is produced, labelled, and ends on the post's open thread or a plain link to it.

Procedure

1. Read the post

Identify: the thesis (the single central claim); the key takeaway; the audience; the tone; length (enough content to thread, or short-form?).

Series awareness: if the post is an episode in a series, the adaptation is the reader's entry point into the whole serial, so use that. The thread's closing link or the newsletter blurb can point at the series ("part 3 of the [X] build — start at part 1: [link]"), and the post's open thread (what's unresolved, what comes next) is natural enticement material: it promises a next episode without manufacturing a cliffhanger. Don't spoil an earlier episode's payoff in the adaptation.

2. Ask what is open

Collect every open question, then send them as one batch in the format of ${CLAUDE_PLUGIN_ROOT}/references/ask-the-author.md (read that file first) and wait for all the answers. If nothing is open, skip this step.

Format(s): if the user specified them, proceed with those. Otherwise add this question:

Which format(s) would you like? Pick any combination. A. Twitter/X thread — opening tweet + numbered thread (up to ~12 tweets) + link close B. LinkedIn post — 150–300 words in the post's own voice, ends on its open thread C. TL;DR — 2–3 sentences: what happened or the claim, what was found, what's still open D. Newsletter blurb — 3–5 sentences teasing the post, ends with [Read more →] E. Other: say what

With no human to answer (a headless run), produce all four, since each is cheap and dropping one is easier for the author than asking again later.

No clear thesis: flag before repurposing, because a fuzzy post produces fuzzy content. Add a question: "A. Sharpen first with /ceh-blog:edit-post B. Proceed as-is C. Other: say what"

Too short to thread (under ~400 words or only 2–3 distinct points): flag it and add a question: "A. LinkedIn or TL;DR instead (recommended) B. Thread anyway, noting where it feels thin C. Other: say what"

3. Produce the output(s)

Produce every requested format, each labelled with a ## heading, in order: Thread → LinkedIn → TL;DR → Newsletter blurb.

Format 1: Twitter/X thread

[Opening tweet]     — the post's opening moment or thought, standalone
[Tweet 2]           — the setup
[Tweets 3–N]        — one beat per tweet, numbered ("2/")
[Closing tweet]     — the open thread, then the link

• Opening tweet: must work standalone — no "A thread on…" filler. Open where the post opens: inside a moment or a thought, in the author's words. Not a teaser line engineered to stop the scroll.
• Numbered tweets ("2/" style from tweet 2): each must make sense on its own — rewrite anything that only makes sense via the previous tweet. No same-sentence-split continuations. Full sentences in the post's register, not staccato one-liners.
• Length: ≤ 280 characters each — count carefully on long ones.
• Count: as many as the post's beats support, usually 5–12, with no filler. If the post only supports 4 good tweets, make 4.
• Closing tweet: the post's open thread in a sentence, then [link]. No follow invite, no reply-bait question, not "That's it!"
• Hashtags: 0–1 total, only if genuinely relevant.

Format 2: LinkedIn post

[Lead]              — the post's opening moment or thought
[Body]              — the story or argument, compressed (2–4 short paragraphs)
[Close]             — the open thread, then [link]

• 150–300 words — one idea developed properly, readable in full.
• Lead inside the moment, not with backstory or a pitch — the first two lines show above "see more", so they carry the post's opening, not "I've been thinking about X lately" and not a lesson stated upfront.
• Match the original post's tone: technical stays technical, personal stays personal, so the writer's voice survives the adaptation.
• Connected paragraphs: 2–4 sentences each, blank line between. No one-line-per-paragraph broetry.
• 0–2 hashtags, only genuinely descriptive ones. In doubt: zero.
• End on the open thread — what's unresolved or what comes next, then [link]. No engagement question.

Format 3: TL;DR

[Sentence 1]    — the thesis / core argument
[Sentence 2]    — the key insight or supporting evidence
[Sentence 3]    — the open thread (what's still unresolved or next)

• 2–3 sentences, no more. A fourth means one of the three is doing too much.
• Must stand alone for someone who never saw the post.
• Active voice, first person where the post is first person: "I found X causes Y", not "Y is caused by X".
• No jargon unless the audience uses it daily.
• Capture the substance, not a summary of it: state what happened or the actual claim, not "This post is about…". No advice to the reader.

Format 4: Newsletter blurb

[Tension or hook]       — a question, problem, or tension the post resolves
[Brief context]         — enough setup to care
[Payoff hint]           — gesture at the insight without giving it away
[Link placeholder]      — [Read more →]

• 3–5 sentences — enough to create intrigue, not enough to replace the click.
• Tension, not summary: pose the question the post answers or name the problem it solves. The reader should finish wanting the answer.
• Don't give away the key insight: if the thesis is "X is the wrong approach", raise the problem X solves and don't reveal that X is wrong. That's the click.
• End with [Read more →].

Rules

Voice. The adaptation carries the post's personal voice onto the platform. It does not switch to platform-native influencer style. If the target repo's CLAUDE.md defines a blog voice, it overrides the format rules above.

• Banned tells apply here too: punchy standalone one-liner paragraphs; aphoristic closers; imperative lessons aimed at the reader ("Don't design your own. Surface theirs."); "If you're building X, then Y" prescriptions; bold pseudo-headers as section labels; tidy meta-takeaway sign-offs; CTA endings ("follow for more", "what do you think?").
• Never invent a scene, feeling, number, or claim the post doesn't contain. The adaptation compresses the post, it doesn't add to it. Reuse the post's own sentences where they fit.
• Endings: each format ends on the post's open thread or a plain link to it, never a lesson and never an engagement prompt. A bare [link] is a link, not a CTA.
• Highly technical post: keep the technical language, don't dumb it down.
• Personal or emotional post: preserve the honest voice, don't make it corporate.
• Post is on disk: read it from the file, and don't overwrite the post.

Output

Single format: output directly with a format label. Multiple: ## heading per format in the order above.

Save each format to its own file, <slug>-<format>.md (thread, linkedin, tldr, newsletter): next to the post when it is on disk, otherwise in .agents_workspace/blog/ with the slug from the post's title in kebab case. Each file fills ${CLAUDE_PLUGIN_ROOT}/assets/repurposed-post.template.md, per ${CLAUDE_PLUGIN_ROOT}/references/repurposed-post-schema.md. A re-run overwrites the format files, never the post. Say where they were saved.

After delivering, ask for feedback as one batch in the same format, one atomic question per adjustment: tone, length, a specific tweet, the LinkedIn opener, each with options (keep, a concrete alternative, Other: say what). Include a question for each thing you'd refine even if they're satisfied. Common candidates: opening tweet reads as summary rather than a moment; LinkedIn opener starts with backstory; a banned tell crept in during compression; blurb gives away too much.

Stop conditions

• No clear thesis → flag it, ask whether to sharpen first or proceed as-is, and wait.
• Too short to thread → flag it and ask whether to switch to LinkedIn or TL;DR instead.