Skip to content
Back to skills

Diataxis Classification

ASecurity

Practical guide for classifying documentation into the four Diataxis quadrants (tutorial, how-to, reference, explanation) — with a decision checklist, anti-pattern catalog, and remediation moves.

  • 7 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added September 23, 2026
ai-agentsrustgoapidocumentation

Works with

  • cli
  • api

Security analysis

A100/100

Scanned September 23, 2026

npx -y skills add mcorbett51090/RavenClaude --skill diataxis-classification --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Diataxis Classification?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Diataxis Classification
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/mcorbett51090-diataxis-classification/badge)](https://www.skillsdirectory.com/skills/mcorbett51090-diataxis-classification)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: diataxis-classification
description: "Practical guide for classifying documentation into the four Diataxis quadrants (tutorial, how-to, reference, explanation) — with a decision checklist, anti-pattern catalog, and remediation moves."
---

# Diataxis Classification

## When to Use This

Before writing a new doc or auditing existing content, to ensure you're writing the right kind of thing — serving the reader's actual need rather than mixing quadrants into docs that help nobody well.

## The Four Quadrants

| Type | Serves | Reader state | Success criterion |
|---|---|---|---|
| **Tutorial** | Learning | Newcomer; wants to be led | Reader completes it and has learned something real |
| **How-to guide** | Accomplishing a task | Practitioner; wants to do X | Reader achieves the specific goal |
| **Reference** | Information lookup | Expert; wants a fact | Reader finds the fact quickly and trusts its accuracy |
| **Explanation** | Understanding | Curious; wants to know why | Reader develops a mental model |

The axes are: **action vs cognition** (doing vs understanding) and **acquisition vs application** (learning vs working).

## Classification Checklist

Ask these questions about your document:

```
1. Is the reader already using the product, or just starting out?
   → Starting out: Tutorial
   → Already using: How-to, Reference, or Explanation

2. Does the reader want to DO something right now?
   → Yes: Tutorial (guided practice) or How-to (autonomous task)
   → No: Reference (lookup) or Explanation (understanding)

3. Is the reader a newcomer being guided, or a practitioner navigating alone?
   → Guided newcomer: Tutorial
   → Autonomous practitioner: How-to

4. Does the reader want a fact (value, parameter, behavior) or a mental model (why, how it works)?
   → Fact: Reference
   → Mental model: Explanation
```

## Anti-Pattern Catalog

| Anti-pattern | Symptom | Remediation |
|---|---|---|
| Tutorial-as-reference | Guided walkthrough ending with a giant parameter table | Split: keep walkthrough as tutorial; extract table to reference |
| How-to buried in tutorial | "Now that you've learned X, here's how to do Y and Z" | Write Y and Z as separate how-to guides; link from tutorial |
| Explanation mixed into reference | API reference with paragraphs about design philosophy | Move philosophy to an explanation doc; reference stays factual |
| Reference used as a tutorial | "See the parameter list to get started" | Write a tutorial that uses the reference; don't merge them |
| Giant catch-all "Overview" page | Contains tutorial steps + design rationale + parameter tables + task instructions | Decompose into one doc per quadrant; overview page links to all four |

## Reference vs How-to Decision

The most commonly confused pair:

| Scenario | Correct type |
|---|---|
| "Here are all the config options and their defaults" | Reference |
| "How to configure authentication with OAuth 2.0" | How-to |
| "Here are all the API endpoints, their parameters, and responses" | Reference |
| "How to authenticate a request" | How-to (uses reference for details) |
| "Here are all the error codes" | Reference |
| "How to handle a 429 rate-limit response" | How-to |

## Writing Signals by Type

**Tutorial signals** — present tense, active voice, guiding tone: "Click Save. You'll see a confirmation…" Minimal explanation, maximum forward momentum.

**How-to signals** — imperative mood, assumes prior knowledge, no hand-holding: "Configure the webhook endpoint." Reader can skip steps they've done before.

**Reference signals** — noun-heavy, consistent structure, scannable: function signatures, tables, parameter lists, exact values. No prose narrative.

**Explanation signals** — discusses trade-offs, uses "because," "in order to," "the reason": "Tokens expire after 15 minutes because long-lived tokens increase the blast radius of a compromise."

## Audit Worksheet

For each existing doc, complete:

```
Doc title: ___________
Intended type (by author): ___________
Actual type (by classification checklist): ___________
Mixed types present: ___________
Remediation: split / rewrite / move / keep
```

## Pitfalls

- Writing tutorials for practitioners — they skip the hand-holding and get lost; they need a how-to.
- Conflating "comprehensive" with "good" — a reference page that explains everything is harder to scan than one that just states facts.
- Writing explanation in the reference page to justify design choices — readers looking up a parameter don't want a philosophy essay; link to the explanation doc.
- Creating an "overview" page to avoid classifying — it becomes the junk drawer that nobody trusts.

## See Also

- [`../../agents/docs-architect.md`](../../agents/docs-architect.md) — Diataxis framework, information architecture, docs-as-code
- [`../../agents/api-reference-writer.md`](../../agents/api-reference-writer.md) — reference and how-to docs for developer APIs

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…