Agent skill

Diataxis

by shepherdjerred in shepherdjerred/monorepo

Author and restructure technical documentation with the Diátaxis framework — tutorials, how-to guides, reference, and explanation.

GPL-3.0Auto-check passedWriting & Content

Install Diataxis

skills CLI
$ npx skills add shepherdjerred/monorepo --skill diataxis -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install shepherdjerred/monorepo diataxis --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/shepherdjerred/monorepo.git skills-src && mkdir -p .claude/skills && cp -r skills-src/packages/dotfiles/dot_agents/skills/diataxis .claude/skills/diataxis && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
diataxis
GitHub stars
112
Token cost
~2.6k tokens
SKILL.md length
1,379 words
Files
1
Skills in repo
63
Repo updated
First seen
Licence
GPL-3.0

At a glance

Author and restructure technical documentation with the Diátaxis framework — tutorials, how-to guides, reference, and explanation.

  • Works in 2 steps: Does it inform action (doing) or… → Does it serve acquisition (the reader is…
  • Reorganising docs
  • SKILL.md covers The compass — use this first, The four kinds, The two distinctions people… and Style rules, plus 4 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

Diataxis is an agent skill from shepherdjerred/monorepo. Author and restructure technical documentation with the Diátaxis framework — tutorials, how-to guides, reference, and explanation. Use when writing or reorganising docs, when a page mixes instruction with description or discussion, when deciding where a piece of content belongs, when a doc "reads badly" without an obvious cause, or when the user mentions Diátaxis, docs structure, or a documentation rewrite.

Its SKILL.md is about 2.6k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Writing & Content, covering Technical writing and Technical documentation. The repository describes itself as: Monorepo for all of my projects. The licence is GPL-3.0.

When your agent uses it

  • Reorganising docs
  • A page mixes instruction with description
  • Deciding where a piece of content belongs
  • A doc reads badly without an obvious cause

Example prompts

  • “reads badly”
  • “/diataxis”

Workflow steps

2 steps, taken from the first numbered list in SKILL.md.

  1. Does it inform action (doing) or cognition (thinking)?
  2. Does it serve acquisition (the reader is at study) or application

What it can do on your machine

Read from SKILL.md and the folder at commit 46ddf2f. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    No scripts in the folder and no shell commands in SKILL.md.

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    Links to these hosts (documentation or services it may open):

    • diataxis.fr

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Diataxis loads about 2.6k tokens when it runs. Until then it costs about 105 tokens; SKILL.md has 1,379 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~105
When it runs · the whole SKILL.md, loaded when a task matches
~2.6k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from shepherdjerred/monorepo at commit 46ddf2f, republished under its GPL-3.0 licence (© shepherdjerred). 1,379 words, ~2,600 tokens.

Download SKILL.mdSave it as .claude/skills/diataxis/SKILL.md (or your agent's skills folder).
name
diataxis
description
Author and restructure technical documentation with the Diátaxis framework — tutorials, how-to guides, reference, and explanation. Use when writing or reorganising docs, when a page mixes instruction with description or discussion, when deciding where a piece of content belongs, when a doc "reads badly" without an obvious cause, or when the user mentions Diátaxis, docs structure, or a documentation rewrite.
user-invocable
true

Diátaxis

Diátaxis says there are exactly four kinds of technical documentation, because there are exactly two axes of any craft. Each kind serves one user need and must not be mixed with the others on one page.

Most bad documentation is not badly written. It is two or three kinds of documentation on one page, each getting in the other's way.

The compass — use this first

When you don't know what a page is, what it should be, or why it feels wrong, ask two questions:

  1. Does it inform action (doing) or cognition (thinking)?
  2. Does it serve acquisition (the reader is at study) or application (the reader is at work)?
InformsServesIt is a
actionacquisitiontutorial
actionapplicationhow-to guide
cognitionapplicationreference
cognitionacquisitionexplanation

Apply it to a whole page, a section, or a single sentence. The most useful move in a rewrite is running the compass over one paragraph at a time and relocating whatever answers a different question than its neighbours.

ActionCognition
Acquisition (at study)TutorialExplanation
Application (at work)How-to guideReference

The four kinds

Tutorial — a lesson

A learning experience. The reader does something meaningful under your guidance and gains confidence. Success is what they can now do, not what they produced.

  • Must not contain: explanation, options, alternatives, or completeness for its own sake.
  • Language: "In this tutorial we will…", "First, do x. Now do y.", "The output should look something like…", "Notice that…"
  • Failure mode: the author cannot resist teaching. Explanation dissolves the learner's attention. One clause is enough — link out for the rest.
  • Show the destination up front. Deliver a visible result at every step. Narrate what to expect and what to notice. Permit repetition. It must work every time.
How-to guide — a recipe

Directions that get a competent reader from a real problem to a result.

  • Must not contain: teaching, background, or reference for completeness.
  • Language: "This guide shows you how to…", "If you want x, do y", "Refer to the x reference for the full list of options."
  • Failure mode: written from the machine's perspective ("To deploy, press Deploy") instead of the reader's goal. Title it for the goal.
  • Address real-world complexity — a guide useful for exactly one narrow case is rarely useful. Practical usability beats completeness. Sequences may fork.
Reference — a description

Facts the reader consults while working. Austere, neutral, authoritative.

  • Must not contain: instruction, discussion, or opinion.
  • Language: "Sub-commands are: a, b, c." "You must not apply b unless c."
  • Failure mode: examples grow into explanation. Illustrate, never instruct.
  • Mirror the structure of the thing described. Adopt consistent patterns — the value of reference is that it is boring and predictable. Tables, not prose.
  • Generated reference (from code, schemas, registries) beats transcribed reference, because it cannot drift.
Explanation — a discussion

Understanding, from a wider angle. The only kind you would read away from the keyboard.

  • Must not contain: steps, or a list of flags.
  • Language: "The reason for x is historically y…", "w is better than z because…", "Some prefer w, which can be good, but…"
  • Failure mode: it absorbs everything nearby, because every topic touches instruction and description. Keep it bounded; link out.
  • Say why: design decisions, constraints, history, alternatives rejected. Opinion and judgement belong here and nowhere else. Make connections.
  • It may be called Concepts, Background, Discussion, or Topics.

The two distinctions people actually get wrong

Tutorial vs how-to is the most common and most damaging conflation. Both are practical, both are step-by-step, so they look alike. The difference is the reader: at study or at work.

TutorialHow-to guide
builds basic competenceassumes competence
a managed path, one line, no choicesforks and branches; the real world
a contrived, safe, repeatable settingwhatever reality throws at it
eliminates the unexpectedprepares for the unexpected
responsibility lies with the teacherresponsibility lies with the user
concrete and particulargeneral, adaptable

It is not the difference between basic and advanced. A how-to can be elementary; a tutorial can be highly advanced.

Reference vs explanation is easier but slips. Rules of thumb: if it is boring and unmemorable it is reference; lists and tables are reference; if you can imagine reading it in the bath it is explanation. The real test is the same one — would the reader reach for this while working, or while reflecting?

Style rules

These are what make documentation read well, independent of structure.

Universal

  • Open with a one-sentence answer to the page's question, then a short summary, then sections. (Wikipedia's lede shape works because it has been refined for 25 years.)
  • One idea per sentence. Under ~25 words. At most one subordinate clause. A sentence with two em-dash asides is three sentences.
  • Bullets are lists of parallel items, not paragraph containers. Cap ~40 words; anything longer wants to be a subsection.
  • Titles state the kind: How to …, … reference, Why … / About ….
  • State a fact once, in one place, and link to it from everywhere else.
  • Navigation lists cap at about seven items. Landing pages introduce their contents in prose — never a bare list of links.
  • No project-management residue: no plan status, no "superseded", no "TODO remains open". That belongs in the tracker, not the docs.
Show full SKILL.md (514 more words)Show less

Working method

Diátaxis is a guide, not a plan. Do not design the perfect structure up front, and never create empty section shells — structure is what emerges from improving pages, not what you impose before writing them.

The loop:

  1. Choose something. Any page, section, or paragraph. Don't go hunting.
  2. Assess it. What need does this serve? How well? Do its language and form match that kind?
  3. Decide one action that improves it now.
  4. Do it, and ship it. Then repeat.

Two techniques worth using:

  • Phantom links. When you feel the urge to explain inside a how-to, link to a reference or explanation page that does not exist yet. The broken links become your work list.
  • Expect it to look worse first. Diátaxis exposes gaps that blur was hiding. That is the tool working, not a setback.

Adaptations and limits

  • The four kinds are a SHOULD, not a MUST. The most common failure in practice is dogmatic application. A link out always beats a digression, but a one-clause aside beats a link the reader must chase.
  • Examples in reference are correct and endorsed — they illustrate. They stop being correct when they start instructing.
  • Cross-link densely. Splitting into four sections without links makes docs harder to use. Every how-to links its reference. Reference links out one-way, so facts keep a single home.
  • Don't add a click to the thing read 95% of the time. Deep-link the highest-traffic pages from the landing page.
  • Troubleshooting is a how-to (How to diagnose …). "Should I use this?" is explanation. Avoid FAQ pages — an FAQ is where content goes when nobody will decide where it belongs.
  • Diátaxis cannot give you accuracy or completeness. It addresses flow, fit, and anticipating the reader. Correctness is still your job.
  • Very large doc sets may need a second axis (subject, audience, platform). That is allowed: Diátaxis is not four boxes. Ask whether the subjects are effectively different products for different readers — if so, lead with subject; otherwise lead with the four kinds.

Starlight layout

The house format for a Diátaxis site in this ecosystem:

text
src/content/docs/
├── index.md          # splash: hero + one section per reader need
├── tutorials/
├── how-to/
├── reference/
└── explanation/      # labelled "Concepts" in the sidebar
  • Files are kebab-case and named for the reader's goal (troubleshoot-notifications.md, not notifications.md).
  • Sidebar groups use { autogenerate: { directory: "…" } }, with order from each page's sidebar.order frontmatter. Autogeneration means a new page is never orphaned.
  • Plain docsSchema(). The directory is the page's kind; do not add a frontmatter field that restates it.
  • The home page routes by need, in this order: Start here / Solve a specific problem / Look something up / Understand how it fits together.
  • .md by default; .mdx only where a page renders generated content.

Page shapes that work:

  • Tutorial: "In this tutorial you will…", a time estimate, a screenshot of the payoff before step 1, numbered ## N. Step sections, "What you did", then "From here" links.
  • How-to: one orienting sentence, numbered steps ordered so each rules out the most ground, tables for facts, asides for traps, "## Related" at the end.

Reference

  • https://diataxis.fr/ — the framework. The most useful page in practice is the compass.
  • In shepherdjerred/monorepo, load monorepo-docs for where documentation belongs, and follow packages/docs/wiki/AGENTS.md for the wiki's rules.

© shepherdjerred, GPL-3.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in packages/dotfiles/dot_agents/skills/diataxis of shepherdjerred/monorepo.

Open the folder on GitHubat commit 46ddf2f

Compare with similar skills

Diataxis next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Diataxis compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Diataxis this skillshepherdjerred/monorepo112—~2.6kAutomated safety check: PassGPL-3.0
Beads Documentation Style Guidegastownhall/beads28k—~3.2kAutomated safety check: PassMIT
Technical Writing Standardcursor/plugins10k10 repos~2.4kAutomated safety check: PassNone
Heym Documentation Articlesheymrun/heym1.4k—~780Automated safety check: PassCustom licence
Developer Docs Technical Writervercel-labs/github-tools131—~3.9kAutomated safety check: PassMIT
Aholo Viewer Docsmanycoretech/aholo-viewer1.1k—~341Automated safety check: PassMIT

Similar skills

  • Sets the house style for the beads user docs: the canonical concept model, required terminology, prose and diagram conventions, and checks before docs work is done.

    28k GitHub stars~3.2k tokensUpdated today
    Writing & ContentAuto-check passed
  • Official

    Applies four layers of technical-writing rules to docs, RFCs, readmes, PR descriptions and commit messages so a tired engineer follows them on the first read.

    10k GitHub starsUsed in 10 repos~2.4k tokens
    Writing & ContentAuto-check passed
  • Creates and updates documentation articles for the Heym platform: category choice, manifest entry, markdown file and cross-links from existing pages.

    1.4k GitHub stars~780 tokensUpdated yesterday
    Writing & ContentAuto-check passed
  • Developer Docs Technical Writer

    vercel-labs/github-tools

    Official

    Writes, reviews and edits developer documentation for SDKs, libraries and frameworks, from getting-started guides and API references to migration guides.

    131 GitHub stars~3.9k tokensUpdated 8 days ago
    Writing & ContentAuto-check passed
  • Aholo Viewer Docs

    manycoretech/aholo-viewer

    Guides writing and maintaining Aholo Viewer documentation: README, AGENTS.md, architecture notes, bilingual manual pages and AI collaboration guides.

    1.1k GitHub stars~341 tokensUpdated 13 days ago
    Writing & ContentAuto-check passed
  • Diataxis

    WebMCP-org/npm-packages

    Write technical documentation following the Diataxis framework by Daniele Procida.

    103 GitHub stars~1.7k tokensUpdated 3 days ago
    Writing & ContentAuto-check passed

More from shepherdjerred/monorepo

All 63 skills in this repo
  • Bun Runtime Best Practices

    shepherdjerred/monorepo

    Bun runtime APIs and current operational patterns for files, processes, modules, networking, databases, tests, and deployment.

    112 GitHub stars~2k tokensUpdated today
    Auto-check passed
  • Bun Test Patterns

    shepherdjerred/monorepo

    Current Bun test runner guidance for discovery, isolation, parallelism, sharding, changed tests, mocks, timers, snapshots, coverage, DOM Testing Library, and integration teardown.

    112 GitHub stars~2k tokensUpdated today
    Auto-check passed
  • Bun Workspaces

    shepherdjerred/monorepo

    Current Bun workspace guidance for isolated and hoisted linkers, catalogs, filters, scripts, dependency classes, lockfiles, lifecycle trust, caches, publishing, TypeScript package exports, and…

    112 GitHub stars~2.2k tokensUpdated today
    Auto-check passed
  • Deep Research

    shepherdjerred/monorepo

    This skill should be used when the user asks to "deep research", "research this topic", "investigate thoroughly", "do a deep dive on", "comprehensive research on", "find everything about", "survey…

    112 GitHub stars~4.7k tokensUpdated today
    Auto-check: notes
  • Figma Use

    shepherdjerred/monorepo

    This skill should be used when the user asks to "create a Figma design", "design in Figma", "make a Figma mockup", "create an app icon", "design UI", "render JSX to Figma", "export from Figma"…

    112 GitHub stars~946 tokensUpdated today
    Auto-check passed
  • Fish Helper

    shepherdjerred/monorepo

    Current Fish shell scripting, functions, abbreviations, completions, variables, events, configuration, plugins, testing, and safety guidance.

    112 GitHub stars~2k tokensUpdated today
    Auto-check passed

Questions about Diataxis

What does Diataxis do?

Author and restructure technical documentation with the Diátaxis framework — tutorials, how-to guides, reference, and explanation. Diataxis is an agent skill from shepherdjerred/monorepo. Author and restructure technical documentation with the Diátaxis framework — tutorials, how-to guides, reference, and explanation.

When should I use Diataxis?

Diataxis fits situations like: reorganising docs; A page mixes instruction with description; deciding where a piece of content belongs; A doc reads badly without an obvious cause.

How do I install Diataxis in Claude Code?

Run `npx skills add shepherdjerred/monorepo --skill diataxis -a claude-code`. Or copy the skill folder (packages/dotfiles/dot_agents/skills/diataxis in shepherdjerred/monorepo) into .claude/skills/diataxis in your project. Claude Code loads it when a task matches its description.

How do I install Diataxis in Codex?

Run `npx skills add shepherdjerred/monorepo --skill diataxis -a codex`. Or copy the skill folder (packages/dotfiles/dot_agents/skills/diataxis in shepherdjerred/monorepo) into .agents/skills/diataxis in your project. Codex loads it when a task matches its description.

Can I use Diataxis in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add shepherdjerred/monorepo --skill diataxis -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/diataxis, .gemini/skills/diataxis, .github/skills/diataxis and .opencode/skills/diataxis in your project.

What does Diataxis need to run?

SKILL.md names no scripts, command-line tools or credentials: Diataxis is instructions for the agent only.

Does Diataxis access the network?

SKILL.md names 1 domain. As links in the text: diataxis.fr. This is read from the text; nothing was executed.

Is Diataxis safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Diataxis use?

Diataxis is published under the GPL-3.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Diataxis use?

About 2.6k tokens (SKILL.md is roughly 10k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Diataxis?

Skills that share tags, products or a category with Diataxis: Beads Documentation Style Guide (gastownhall/beads, 28k stars), Technical Writing Standard (cursor/plugins, 10k stars), Heym Documentation Articles (heymrun/heym, 1.4k stars) and Developer Docs Technical Writer (vercel-labs/github-tools, 131 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Diataxis?

shepherdjerred (a GitHub user) maintains it in shepherdjerred/monorepo, which has 112 GitHub stars. The repository holds 63 skills in this directory. The repository was last updated on October 7, 2026.

Source: shepherdjerred/monorepo on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.