effective-claude-code-usage
AGENTS.md1 starsChanged 11 days agoCC0-1.0View on GitHub
A single-page Quarto presentation rendered to RevealJS slides and deployed as a static site via GitHub Pages. _site/, _extensions/, .quarto/ and .venv/ are gitignored; never commit them. - Single-file deck. All slides live in index.qmd; do not split it. Slides are separated by ## headings. Use Quarto's RevealJS dialect: fenced divs (:::), columns (.columns / .column), raw HTML blocks ({=html}), and the {.smaller} class for dense slides. Follow the layout patterns of neighbouring slides. - Theme and colours. Keep the…
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,542 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository7 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- llms-full.txtllms.txt
- llms.txtllms.txt
- .well-known/llms-full.txtllms.txt
- .well-known/llms.txtllms.txt
- check-accessibilitySkill
- generate-diagramsSkill
The file
We split this AGENTS.md into the rules it gives an agent
6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 42
Single-file deck: All slides live in index.qmd.
- Very specific: line 47
Theme and colours: Keep the theme black and the format revealjs.
- Distinctive: line 53
Diagrams and text must not restate each other: If a diagram already shows a concept, do not repeat its labels as cards or bullets below it.
- Distinctive: line 61
Sources: Every factual claim has a source citation at the bottom of its slide in a small-font centred div.
- Distinctive: line 63
Spelling and punctuation: Use British spelling in prose (colour, licence, catalogue, artefact) and the Oxford comma in lists of three or …
- Very specific: line 84
Discoverability files: Update meta-tags.html when the title, description, or social card changes, and llms.txt / llms-full.txt (and their …
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md 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
AGENTS.md
A single-page Quarto presentation rendered to RevealJS slides and deployed as a static site via GitHub Pages.
Repository layout
index.qmd # All slide content (the only file you usually need to edit)
_quarto.yml # Quarto project config (output dir, resources list)
_quarto-a11y.yml # Opt-in profile enabling the axe accessibility checker (`just axe`)
accessibility.html # Compatibility fixes supplementing the a11y extension
style.css # Custom RevealJS theme (fonts, colours, component classes)
meta-tags.html # OpenGraph, Twitter Card, JSON-LD, and analytics tags
justfile # Command runner
media/ # Generated diagrams (and their script), screenshots, social card
llms.txt, llms-full.txt # Machine-readable summaries (mirrored in .well-known/)
robots.txt, sitemap.xml # Static crawl files; rarely change
pyproject.toml, uv.lock # Python dependencies for Quarto's Jupyter engine (uv)
CLAUDE.md # Imports this file for Claude Code
LICENSE # CC0 dedication_site/, _extensions/, .quarto/ and .venv/ are gitignored; never commit
them.
Commands
just install # Install Python dependencies and Quarto extensions
just update # Update Python dependencies
just render # Render index.qmd to _site/ (run after every change)
just preview # Live-reload dev server
just clean # Remove build artefacts
just check # Verify Quarto and Python setup
just axe # Preview with the axe accessibility checker enabledKey conventions
- Single-file deck. All slides live in
index.qmd; do not split it. Slides are separated by##headings. Use Quarto's RevealJS dialect: fenced divs (:::), columns (.columns/.column), raw HTML blocks ({=html}), and the{.smaller}class for dense slides. Follow the layout patterns of neighbouring slides. - Theme and colours. Keep the theme
blackand the formatrevealjs; restyle throughstyle.css. Lay out content with the classes instyle.cssrather than inline styles:.card-green,.card-blue,.card-amber,.card-purple,.card-red,.card-subtle(plus.card-compact), and.terminal-boxfor terminal output. New widgets follow the naming patterns already instyle.css. - Diagrams and text must not restate each other. If a diagram already shows a concept, do not repeat its labels as cards or bullets below it. Text below a diagram is for what the diagram does not show: quantitative data, actionable guidance, caveats, or decision criteria. If removing the text loses no information, remove it.
- Generated diagrams.
media/generate_diagrams.pyproduces the WebP diagrams with matplotlib and the Caveat font. Load thegenerate-diagramsskill before adding or changing one. - Sources. Every factual claim has a source citation at the bottom of its slide in a small-font centred div. Keep this pattern.
- Spelling and punctuation. Use British spelling in prose (colour, licence,
catalogue, artefact) and the Oxford comma in lists of three or more. Leave
code, identifiers, file names, URLs, quotations, and proper names (
licensein YAML,.well-known/api-catalog) as they are. - Spoken notes. The read-aloud control speaks the
.notescontent. Write complete, audience-facing narration; keep staging cues, private reminders, and delivery checklists outside these blocks. - Accessibility. Images must have
fig-alttext; raw HTML widgets userole="img"andaria-label. Load thecheck-accessibilityskill before changing markup, styles, widgets or the a11y setup. - Icons. Icons are HTML spans backed by only the required SVG path data in
style.css; no icon font or Quarto icon extension. When adding one, add only its mask data, preserve the source licence attribution, and keep an accessible label where the icon conveys meaning. - No code execution. The front matter sets
execute: eval: false; code blocks are display-only. Do not enableeval: trueunless the deck genuinely needs computed output. - Python dependencies. The deck declares
jupyter: python3; the virtualenv only satisfies Quarto's engine. Depend only onipykernel,nbclientandpyyaml. Do not add thejupytermetapackage: it pulls in JupyterLab and Notebook and only attracts irrelevant security alerts. - Discoverability files. Update
meta-tags.htmlwhen the title, description, or social card changes, andllms.txt/llms-full.txt(and their.well-known/mirrors) when the content changes significantly. - No new top-level files without a clear reason; the structure is flat.
CI/CD
.github/workflows/build-presentation.yamlcalls a reusable workflow fromIndrajeetPatil/workflows: pushes tomainbuild and deploy to GitHub Pages, pull requests build without deploying. Do not inline or modify that workflow here.- Reference reusable workflows as
@main, never a commit SHA: they are first-party, and a pinned SHA once went five months stale. - Dependabot updates GitHub Actions and
uvdependencies weekly. check-link-rot.yamlchecks external links on pushes, pull requests and weekly (Sunday 00:00 UTC); list false positives in.lycheeignore.
Agent skills
Skills live in .agents/skills/<name>/SKILL.md. Load one before its task:
| Skill | Load before |
check-accessibility | Changing slide markup, styles, widgets, menus or the a11y setup; any accessibility audit |
generate-diagrams | Adding or changing a diagram in media/generate_diagrams.py |
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.

