content-atomisation
Skill0 starsChanged 31 days agoMITView on GitHub
Turns one long piece of content (a blog post, guide, report, talk transcript or podcast) into a sequenced, channel-by-channel plan of standalone derivative pieces, each with one claim, one link back and one call to action, then drafts them on request. Use when someone says "repurpose this", "atomise this post", "turn this into social posts", "content plan from this article", "break this into channel pieces", "derivative content", "what can I get out of this one piece", or asks to check or fix an existing repurposing plan, content map or derivative schedule before it is scheduled.
In this repository4 other files it gives its agents.
- ai-search-visibilitySkill
- keyword-researchSkill
- linkedin-postSkill
- seo-schema-gateSkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 7 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Standalone: Every derivative teaches one thing and makes sense on its own.
- Distinctive: line 21
If two rows say the same thing in different words, cut one.
- Distinctive: line 22
No padding: If the source has one real point, the map has one derivative.
- Distinctive: line 39
The distinct points: every separate claim, number, example, story, framework or objection the piece contains.
- Very specific: line 66
Default pacing, from references/sequencing.md: social 1 day after the hub, email 3 to 5, short video 5 to 7, carousel 7 to 10, audio 14 to …
- Distinctive: line 72
If the person hands you an existing map to check, start here: run the checks below on their map and report each problem by row, including …
- Very specific: line 75
If you can run scripts, write the map as TSV (fields in the order sequence, lag_days, channel, format, core_claim, link_target, cta) and …
- Distinctive: line 86
If you cannot run scripts, check the same seven things by reading the table.
- Distinctive: line 105
If the person gives a start date, add actual dates.
- Very specific: line 114
Use the source's facts only.
- Very specific: line 117
Keep the person's voice. If they have the linkedin-post skill from this plugin, use it for LinkedIn rows so the voice is learned from their …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Content atomisation
One good long piece can feed weeks of channels. Most repurposing fails in one of two ways: the same point reworded six times, or fragments that only make sense if you read the original. This skill avoids both by planning first and drafting second.
The output is an atomisation map: one row per piece, each with a distinct angle, one core claim, one link and one call to action, in a publish order that never sends anyone to a page that is not live yet.
The four rules
- Standalone. Every derivative teaches one thing and makes sense on its own. No "part 1 of 3", no "as I said last week", no teaser that holds back the payoff.
- Distinct angles, not paraphrases. Each piece answers a different reader question. If two rows say the same thing in different words, cut one.
- No padding. If the source has one real point, the map has one derivative. Say so. Supporting lines (the evidence, a restatement, the rule of thumb that follows from it) are not extra points: a 100-word opinion piece gets the hub plus one or two pieces, even when the person asks for a post a day. A 600-word opinion piece might support one or two; a 3,000-word guide might support eight.
- Hub first. The long-form piece (the hub) publishes before any derivative, and every derivative links back to it or to a page that links to it.
Step 1: read the source
Accept pasted text, a file, or a URL (fetch it). Extract and show the person:
- Title
- Core thesis in one sentence
- Hub URL, or a placeholder slug if it is not live yet
- Approximate length
- The distinct points: every separate claim, number, example, story, framework or objection the piece contains. This list is the raw material. Number them.
Then ask one question if you need to: which channels do they actually use? Do not plan for channels they do not run. If they do not say, default to the hub plus LinkedIn and email.
Step 2: build the map
One row per piece. Read references/channel-formats.md for what each channel needs and
references/linking-and-ctas.md for link and CTA rules.
| Field | Rule |
sequence | 1 is the hub. Derivatives start at 2, in publish order. |
lag_days | Days after the hub publishes, always a number. Hub is 0. Derivatives 1 or more, never going backwards. For a channel on a fixed schedule (a monthly newsletter), use the first send on or after the pacing default and say which send you assumed; never write "TBC" or "next send". |
channel | Where it goes: linkedin, x, email, newsletter, short-video, carousel, podcast, blog. |
format | single post, thread, teaser, talking head, 8-slide carousel, and so on. |
angle | The reader question this piece answers. Must differ from every other row. |
core_claim | One sentence. The single claim this piece makes. Two facts joined with "and" are two claims: pick one. |
link_target | The one URL this piece sends people to, written out in full: the hub, or a related page that links to the hub. |
cta | One specific action, aligned with the link. |
Match each distinct point from Step 1 to the channel where it works best: a surprising number suits a short post, a step-by-step suits a carousel or email, a story suits video or audio. Not every point needs a piece.
Default pacing, from references/sequencing.md: social 1 day after the hub, email 3 to 5,
short video 5 to 7, carousel 7 to 10, audio 14 to 21, and no more than one piece per
channel per week from the same source.
Step 3: validate the map
If the person hands you an existing map to check, start here: run the checks below on their map and report each problem by row, including any derivative scheduled on the hub's own day.
If you can run scripts, write the map as TSV (fields in the order sequence, lag_days,
channel, format, core_claim, link_target, cta) and run:
bash scripts/validate_map.sh map.tsvIt checks: one hub at day 0, at least one derivative, no empty fields, sequences 1..N with no gaps, no duplicate claims, and no derivative scheduled before the one before it. Works with the stock bash and awk on macOS and Linux.
If you cannot run scripts, check the same seven things by reading the table. Then do the check the script cannot: read the angles side by side and cut any that are paraphrases.
Step 4: present the plan
# Atomisation plan: <title>
**Hub:** <url> | **Thesis:** <one sentence> | **Pieces:** <N> (1 hub + <N-1> derivatives)
| # | Day | Channel | Format | Angle | Core claim | Link | CTA |
| --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | 0 | blog | guide | - | <thesis> | <hub> | N/A |
| 2 | 1 | linkedin | single post | ... | ... | <hub> | ... |
**Left out:** <points from the source not used, and why>
**Before publishing:** hub must be live on day 0; check every link resolves.If the person gives a start date, add actual dates. If not, keep day offsets.
Step 5: draft (when asked)
Draft the pieces the person wants, one per section, each labelled with its row number. For every draft:
- Open with the core claim or the most specific detail, not a preamble.
- Stay within the channel's norms in
references/channel-formats.md. - Use the source's facts only. Never add a statistic, quote or example that is not in the
source; if a piece needs one, mark
[ADD: ...]. - End with the row's CTA. One.
- Keep the person's voice. If they have the
linkedin-postskill from this plugin, use it for LinkedIn rows so the voice is learned from their own posts.
For visual formats (carousel, video), draft the slide text or script and a one-line visual note per slide or scene.
Example
User: Repurpose this 2,200-word post on why small venues lose bookings to slow email replies. We use LinkedIn, a monthly newsletter and Instagram.
Claude: Thesis: venues that reply to enquiries within an hour win roughly twice as many bookings as those replying next day. I found five distinct points; two are strong enough for their own pieces, one suits a carousel, two I would leave out because they repeat the thesis. Plan:
# Day Channel Angle Core claim 1 0 blog - Reply speed decides most venue bookings 2 1 "Does reply time really matter?" The one-hour reply figure from the post's own data 3 7 instagram carousel "How do I reply faster without hiring?" Four templates that cover most enquiries 4 12 newsletter "What do fast venues do differently?" One venue's before and after Want me to draft rows 2 to 4?
Configuration
Two things vary per person:
- Channels and cadence. Which channels they run and how often, so the plan fits their real schedule.
- Link targets. Their hub URL and any related pages a derivative could point to instead. If they have an internal-linking map, use it; otherwise link to the hub.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

