Skip to content
Back to search

content-atomizer

MarketplaceWorks with Claude CodeGitHub

One article or transcript into native posts per platform, each with its own angle.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/mavrin-xv--content-atomizer/badge.svg)](/listings/mavrin-xv--content-atomizer)

HTML

<a href="/listings/mavrin-xv--content-atomizer"><img src="/api/listings/mavrin-xv--content-atomizer/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

1. Install, then open Claude Code in your project. 2. Paste your source and ask:

   > Here is our article on how 14 clients cut their month-end close from
     9 days to 3. Atomize it: a thread, a Telegram post, a carousel and a
     TikTok script. We do not post on LinkedIn.

3. You get a 3 to 5 line plan to approve, then `atoms/<slug>/pieces.md`, `slides.json` and the rendered slides. Every file is checked by its script before you see it.

Run the scripts on the invented example first, from the repo root:

python3 skills/atomize-plan/scripts/plan_check.py examples/northwind-books/plan.json
python3 skills/platform-writer/scripts/post_check.py examples/northwind-books/pieces.md --source examples/northwind-books/source.md
python3 skills/carousel-builder/scripts/render_carousel.py examples/northwind-books/slides.json -o /tmp/carousel.html --png /tmp/slides
python3 skills/hook-library/scripts/hook_bank.py score "14 bookkeeping clients used to close 9 days late. Now it takes 3."

Sample output:

PASS 6 atoms, each with evidence found in the source
PASS linkedin skipped: not an active account for this business
PASS platform-facts fresh until 2027-02-18

PASS x: thread: within 280 chars per post
PASS tiktok: video-script: hook lands in 3s with 7 words
PASS tiktok: video-script: about 29s spoken (73 words)
INFO highest phrase overlap between two pieces: 5%

8/10  14 bookkeeping clients used to close 9 days late. Now it takes 3.
  +2 has a number
  +1 length 13 words
  type: time-transformation

Two of the eight rendered slides:

The whole run for an invented bookkeeping firm (source, plan, four pieces, slides, hooks, comments and drafted replies) is in [examples/northwind-books/](examples/northwind-books/).

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 15 lines
  1. {
  2. "name": "content-atomizer",
  3. "plugins": [
  4. {
  5. "name": "content-atomizer",
  6. "source": "./",
  7. "description": "Turns one article, transcript or talk into native posts per platform: an X thread, a Telegram or newsletter post, rendered carousels and a short video script, each with its own angle. Plans from evidence in the source, checks limits, links, hooks and invented numbers with scripts, builds a hooks library from your own results and triages the replies.",
  8. "version": "0.1.0"
  9. }
  10. ],
  11. "owner": {
  12. "name": "Artem Mavrin",
  13. "url": "https://github.com/mavrin-xv"
  14. }
  15. }
README.md · README · 156 lines
  1. ![content-atomizer](assets/banner.png)
  2. # content-atomizer
  3. **Turn one article, transcript or talk into native posts per platform: an X thread, a Telegram or newsletter post, rendered carousels and a short video script, each with its own angle. For creators, consultants and small teams who write one good piece a week and want it to reach every feed they use.**
  4. Most repurposing pastes the same paragraph into five apps, invents a number on the way, and puts a link where the platform buries it. This pack plans from what the source actually says, writes each piece for its room, and checks limits, links, hooks and every number with a script.
  5. ![License: MIT](https://img.shields.io/badge/license-MIT-blue) ![Works with Claude Code](https://img.shields.io/badge/works%20with-Claude%20Code-orange) ![Python 3.8+ stdlib](https://img.shields.io/badge/python-3.8%2B%20stdlib-green)
  6. ```bash
  7. git clone https://github.com/mavrin-xv/content-atomizer && cd content-atomizer
  8. ./install.sh # or: ./install.sh --project (this project only)
  9. ```
  10. <!-- install-notes:start -->
  11. Other ways in: as a Claude Code plugin, run `/plugin marketplace add mavrin-xv/content-atomizer` and then `/plugin install content-atomizer@content-atomizer`. For Codex or another agent that reads `SKILL.md`, copy the folders under `skills/` into its skills directory. On Windows, run `install.sh` from Git Bash or WSL and type `py -3` where these docs say `python3`.
  12. <!-- install-notes:end -->
  13. ## What it solves
  14. | Pain | What the pack does |
  15. |---|---|
  16. | Every platform gets the same text | `atomize-plan` pulls the ideas out of the source first, then gives each platform its own treatment and hook type, or skips it with a reason |
  17. | Posts quietly grow numbers the source never had | Every atom carries an exact quote from the source, and `post_check.py` fails any digit in a post that is not in the source |
  18. | Threads die because the link is in the first post | `post_check.py` fails a link in an X body, a clickable link promised on Instagram or TikTok, a spoken hook that lands after 3 seconds and posts over the limit |
  19. | Carousels with a paragraph on slide 1 and no ask on the last | `carousel-builder` has five frameworks, a linter for slide rules, a renderer to PNG and PDF, and a visual QA loop with a blind read |
  20. | You guess which hooks work | `hook-library` keeps your first lines with one metric each and ranks the frames you used three or more times by median result |
  21. | The first hour of comments goes to "great post" | `reply-triage` ranks comments into six tiers, flags price and refund questions for a person, and checks the drafts |
  22. ## What's inside
  23. | Skill | What it does | Say |
  24. |---|---|---|
  25. | [atomize-plan](skills/atomize-plan/SKILL.md) | Core insight, atoms with evidence, a treatment per platform, budget and freshness checks | "atomize this", "repurpose this article" |
  26. | [platform-writer](skills/platform-writer/SKILL.md) | X thread, Telegram or newsletter post, Threads, LinkedIn, captions, short video scripts, checked and scored | "write the thread", "TikTok script from this" |
  27. | [carousel-builder](skills/carousel-builder/SKILL.md) | Five frameworks, slides.json, lint, render to HTML, PNG or PDF, visual QA | "make a carousel", "LinkedIn document post" |
  28. | [hook-library](skills/hook-library/SKILL.md) | A CSV of your hooks with nine types, slot patterns, scoring and ranking by results | "build a hook library", "score this hook" |
  29. | [reply-triage](skills/reply-triage/SKILL.md) | Six-tier comment ranking, needs-a-person flags, draft checks | "help me reply to comments" |
  30. Scripts (Python standard library, each with `--help`): `plan_check.py`, `post_check.py`, `carousel_lint.py`, `render_carousel.py`, `hook_bank.py`, `triage_replies.py`. PNG and PDF export uses Playwright if you have it.
  31. ## Quick start
  32. 1. Install, then open Claude Code in your project.
  33. 2. Paste your source and ask:
  34. ```
  35. > Here is our article on how 14 clients cut their month-end close from
  36. 9 days to 3. Atomize it: a thread, a Telegram post, a carousel and a
  37. TikTok script. We do not post on LinkedIn.
  38. ```
  39. 3. You get a 3 to 5 line plan to approve, then `atoms/<slug>/pieces.md`, `slides.json` and the rendered slides. Every file is checked by its script before you see it.
  40. Run the scripts on the invented example first, from the repo root:
  41. ```bash
  42. python3 skills/atomize-plan/scripts/plan_check.py examples/northwind-books/plan.json
  43. python3 skills/platform-writer/scripts/post_check.py examples/northwind-books/pieces.md --source examples/northwind-books/source.md
  44. python3 skills/carousel-builder/scripts/render_carousel.py examples/northwind-books/slides.json -o /tmp/carousel.html --png /tmp/slides
  45. python3 skills/hook-library/scripts/hook_bank.py score "14 bookkeeping clients used to close 9 days late. Now it takes 3."
  46. ```
  47. Sample output:
  48. ```
  49. PASS 6 atoms, each with evidence found in the source
  50. PASS linkedin skipped: not an active account for this business
  51. PASS platform-facts fresh until 2027-02-18
  52. PASS x: thread: within 280 chars per post
  53. PASS tiktok: video-script: hook lands in 3s with 7 words
  54. PASS tiktok: video-script: about 29s spoken (73 words)
  55. INFO highest phrase overlap between two pieces: 5%
  56. 8/10 14 bookkeeping clients used to close 9 days late. Now it takes 3.
  57. +2 has a number
  58. +1 length 13 words
  59. type: time-transformation
  60. ```
  61. Two of the eight rendered slides:
  62. <img src="docs/img/slide-01.png" width="260" alt="Hook slide: Close the month in 3 days, not 9"> <img src="docs/img/slide-04.png" width="260" alt="Hack slide: one receipts inbox">
  63. The whole run for an invented bookkeeping firm (source, plan, four pieces, slides, hooks, comments and drafted replies) is in [examples/northwind-books/](examples/northwind-books/).
  64. ## How it works
  65. ```
  66. source.md (article / transcript / talk)
  67. |
  68. atomize-plan + plan_check.py ---------> plan.json (atoms with evidence,
  69. | a treatment per platform)
  70. +--------------------+
  71. v v
  72. platform-writer carousel-builder
  73. + post_check.py + carousel_lint.py + render_carousel.py
  74. | |
  75. pieces.md slides.json --> carousel.html, PNG, PDF
  76. (thread, Telegram,
  77. captions, video script)
  78. |
  79. v you post
  80. comments.txt --> reply-triage + triage_replies.py --> replies.md
  81. results --> hook-library + hook_bank.py --> hooks.csv
  82. (feeds the hook type for the next plan)
  83. ```
  84. The model makes the judgement calls: what the point is, which idea fits which feed, every word. The scripts do the counting, so a 281-character post, a link in the wrong place or a number the source never had is caught every time. Longer walkthrough: [docs/how-it-works.md](docs/how-it-works.md). File formats: [docs/file-formats.md](docs/file-formats.md).
  85. ## Configure
  86. | What | Where |
  87. |---|---|
  88. | Your accounts, voice and asks | `templates/brief.md`, filled once |
  89. | Platform limits and link rules | the `LIMITS` and `NO_LINK` tables at the top of `post_check.py` |
  90. | Ranking facts and their verify date | `skills/atomize-plan/references/platform-facts.md` (`last_verified`, `ttl_days`) |
  91. | Carousel look | `render_carousel.py --theme paper --accent "#hex"` (themes: dark, light, editorial, paper), or `theme` and `accent` in slides.json |
  92. | Video length | `post_check.py --max-seconds 45` |
  93. | Hook metric per platform | your choice, one per platform; see `skills/hook-library/references/collecting.md` |
  94. No API keys. Nothing in the pack calls a network service.
  95. ## Honest limits
  96. - The scripts were run on the invented example and on broken copies of it. The pack has not been measured against real posting results; it does not promise reach.
  97. - Ranking facts are dated hypotheses. The X weights come from the 2023 open-source release and may have changed. `plan_check.py` fails when the facts file passes its verify date, so someone has to re-check it.
  98. - The source check catches digits, not spelled-out numbers, and not claims without numbers. The model and you still check those.
  99. - Hook type guessing and comment tiers are keyword rules. They misread sarcasm, jokes and other languages; correct them by hand.
  100. - Text and carousels only. Video scripts are written and timed, not produced. There is no scheduling, posting or analytics fetching.
  101. - The carousel themes are deliberately plain. The linter cannot see a cramped slide; the visual QA loop is done by eye.
  102. ## Credits
  103. - Four treatments, the fit table, the 2-3 platform budget, the five-dimension quality gate, reply tiers and the dated platform facts: inspired by `content-atomizer` in https://github.com/boringmarketer/boring-marketing-skills (MIT).
  104. - Content pyramid and per-platform action items: inspired by `content-repurposing` in https://github.com/openclaudia/openclaudia-skills (MIT).
  105. - No duplicate hook patterns across platforms and the zero-context fallback: inspired by `content-atomizer` in https://github.com/moizibnyousaf/marketing-cli (MIT).
  106. - Atoms before writing and the analyze-only mode: inspired by https://github.com/agricidaniel/claude-repurpose (MIT).
  107. - Thin sources get fewer pieces; atomize ideas, not paragraphs: inspired by `content-atomizer` in https://github.com/sanky369/vibe-building-skills (MIT).
  108. - Intro, content and outro slides and slide data as JSON: inspired by https://github.com/FranciscoMoretti/carousel-generator (MIT). No code copied.
  109. - Engagement weights: `projects/home/recap/README.md` in https://github.com/twitter/the-algorithm-ml (AGPL-3.0, numbers cited as facts, no code used).
  110. - Repo and skill layout: modelled on https://github.com/anthropics/skills and https://github.com/obra/superpowers (MIT).
  111. ## Contributing
  112. See [CONTRIBUTING.md](CONTRIBUTING.md). Agents working on the repo: [AGENTS.md](AGENTS.md).
  113. <!-- author:start -->
  114. ## Author
  115. Built and maintained by [Artem Mavrin](https://github.com/mavrin-xv). This repo is one of a set of agent skill packs for marketing and operations work. Each one installs on its own; the full list is in [agent-skill-packs](https://github.com/mavrin-xv/agent-skill-packs). If a pack saved you an afternoon, a star helps other people find it.
  116. <!-- author:end -->
  117. ## License
  118. MIT. See [LICENSE](LICENSE).

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.