read-not-decorate
Skill1 starsChanged 5 months agoMITView on GitHub
Use when someone hands you a finished long-form markdown document (a report, an essay, an archived article, a research write-up) and wants it turned into a single self-contained HTML file meant to be read or kept — not a slide deck, not a marketing page, not a redesign. Trigger on requests like "typeset this report as HTML", "make a clean reading page from this markdown", "archive this article as a standalone HTML I can keep", or "turn this write-up into a page I can send to a client". The skill assembles; it does not rewrite the author's words, add arguments, or add decoration.
In this repository9 other files it gives its agents.
- audit-the-hustleSkill
- backtest-your-frameworkSkill
- distill-not-transcribeSkill
- extract-narrative-methodsSkill
- handoff-disciplineSkill
- missing-not-zeroSkill
- report-deck-zhSkill
- stress-test-the-planSkill
- strip-the-ai-voiceSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 7 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
Before laying out anything, pick one.
- Distinctive: line 41
If the request doesn't say, default to archival and state that you did.
- Very specific: line 58
Every file here has a Chinese twin (*.zh.md, demo.zh.html) with identical rules.
- Distinctive: line 63
Read the whole document first.
- Distinctive: line 70
Wait for the go-ahead or adjustments.
- Distinctive: line 76
Assemble, never author: Not one of the author's words changes.
- Distinctive: line 77
Components are called by structure, not to fill space: No structure, no component — a plain paragraph is the correct default.
- Distinctive: line 85
No engagement furniture: No share/like buttons, no reading-time estimate, no "if you found this useful" closers, no emoji, no stock images, …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Anti-Decoration Documents
Turn a finished long-form markdown document into one self-contained HTML file that gets read, not decorated.
What this is, in one line
You are an assembler, not a writer. The author's text is fixed: you never rewrite a sentence, never "improve" phrasing, never add a point. You choose a reading mode, lay the text out under a strict visual system, and call a component only where the text already has that structure. Everything that fights reading — color, shadow, stock imagery, engagement furniture — is removed on purpose.
Decide the mode first (this changes everything downstream)
Before laying out anything, pick one. They serve opposite readers and the choice decides which components survive.
- Archival — the reader is the author, months from now. Keep the reasoning, the basis for each judgment, the uncertainty that was true at the time. Optimized for re-scanning, not for flow.
- Distribution — the reader is a client or stranger with none of the author's context. Keep clean, independently-standing conclusions and credible reasons. Strip internal revision history and confidence markers.
If the request doesn't say, default to archival and state that you did.
The map (load on demand, do not read all of this up front)
This file is the index. Pull the right reference file when you hit that part of the job — not before.
| When you are… | Load |
| Setting type, color, spacing, the page skeleton | reference/visual-system.md |
| Deciding which structural component a passage needs | reference/components.md |
| Choosing archival vs distribution, and what each drops | reference/modes.md |
| Writing any text you generate (labels, nav, summaries) | reference/copy-restraint.md |
A worked output lives at examples/demo.html — it is a document built with this
skill, about this skill. Open it to see the target.
Every file here has a Chinese twin (*.zh.md, demo.zh.html) with identical
rules. Work in whichever language the source document is in.
Workflow
- Read the whole document first. All of it.
- Output a short confirmation (no half-finished HTML): the chosen mode, the skeleton (H1 + every H2, numbered), and the component trigger points (which passage maps to which component, one line each). If a structured passage fits no listed component, list it as a candidate component and ask — do not force it into the nearest one and do not flatten it into a plain paragraph.
- Wait for the go-ahead or adjustments.
- Produce the complete single
.htmlfile. - Report in ≤5 lines: file path, section count, components used, file size.
Hard rules (these are the product)
- Assemble, never author. Not one of the author's words changes.
- Components are called by structure, not to fill space. No structure, no component — a plain paragraph is the correct default. Bias to fewer.
- One accent color. Everything else earns attention through whitespace and type weight, never through a second color.
- Self-contained. Inline all CSS/JS, system-font stack only, no network, no framework, no CDN. Double-clicking the file must work offline forever.
- Delete every adjective the sentence survives without. Applies only to text you write (nav, labels, summaries) — never to the author's body text.
- No engagement furniture. No share/like buttons, no reading-time estimate, no "if you found this useful" closers, no emoji, no stock images, no AI-generated illustration (any inline SVG you draw is thin and linear).
What this skill is not for
Short notes, unfinished fragments, raw chat logs, slide decks, dashboards, or anything that needs a redesign of the content rather than a layout of it. If the input isn't a finished long-form document, say so and stop — don't HTML it for the sake of it.
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.

