diataxis-writer
Write, restructure, classify, and review documentation with the Diataxis framework.
Skill by tronghieu
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/tronghieu-diataxis-writer
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents, backend, security, writing
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Shell scripts
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/diataxis-writer
What's inside
- Skill name
diataxis-writer- Description
- Write, restructure, classify, and review documentation with the Diataxis framework. Use when the user asks to create, improve, audit, or reorganize documentation of any kind: knowledge bases, help centers, onboarding guides, process docs, manuals, runbooks, product or API docs, tutorials, how-to guides, references, explanations. Also use when the user mentions Diataxis, docs taxonomy, or documentation architecture, or asks why a document feels confusing. Also use for Simplified Technical English (ASD-STE100) or controlled-language rewrites of English technical docs. Not for marketing, legal, sales, press, or fiction writing unless the user wants a Diataxis-style analysis.
- License
- MIT
- Instructions
- 261 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (9)
- README.md
- README.vi.md
- README.zh.md
- SKILL.md
- evals.json · evals/
- ai-prose-tells.md · references/
- diataxis-patterns.md · references/
- ste-rules.md · references/
- classify-doc.sh · scripts/
SKILL.md (first lines)
# Diataxis Writer
Use Diataxis to align documentation with the reader's actual need. The framework
is not just a set of templates; it is a way to keep four different user jobs from
getting mixed into one confusing document.
## Explaining Diataxis to Users
When the user asks what Diataxis is, why it matters, or why this framework is a
good fit, explain it briefly before applying it:
Diataxis is a documentation framework that separates docs by the reader's mode of
work: learning, doing, looking up facts, or understanding concepts. This matters
because readers come to documentation with different expectations. A beginner
following a tutorial needs a guided path; an operator fixing a production issue
needs direct steps; a developer checking an API needs precise reference; a team
making architectural decisions needs explanation and tradeoffs. Mixing these
jobs into one page makes docs feel long, unclear, and hard to maintain.
Use these benefits when recommending Diataxis:
- It clarifies what each document is for before writing begins.
- It reduces bloated "everything pages" by separating learning, task execution,
lookup, and conceptual understanding.
- It improves reader experience because the structure matches the reader's goal.
- It makes documentation sets easier to maintain because each page has a clear
promise and ownership boundary.
- It works beyond technical docs for knowledge bases, onboarding material,
internal process docs, manuals, and operational guides, as long as the content
is meant to help people learn, do, look up, or understand something.
Do not oversell Diataxis as universal. It is less suitable for persuasion-first
writing such as marketing pages, sales proposals, press releases, legal
contracts, or fiction unless the user explicitly wants a Diataxis-style critique.
## Core Model
Classify the document by the reader's intent:
| Reader need | Diataxis type | Document promise |
| --- | --- | --- |
| Learn by doing | Tutorial | "Follow this path and you will gain basic competence." |
| Complete a task | How-to guide | "Follow these steps to achieve this outcome." |
| Look up facts | Reference | "Find accurate, complete information quickly." |
| Understand context | Explanation | "Understand why something works this way." |
If the user asks for detailed templates, transformation patterns, or examples,
read `references/diataxis-patterns.md`.
## Workflow
1. **Identify the reader's job.** Decide whether the reader is trying to learn,
do, look up, or understand. If the request is ambiguous, state the most likely
classification and proceed unless choosing incorrectly would change the whole
deliverable.
2. **Define audience and scope.** Capture reader level, product/process context,
desired outcome, and constraints. Keep these visible while writing.
3. **Choose one primary document type.** Let the primary type control structure,
tone, and depth. Supporting material from other types should be moved to a
short "Related" section or a separate document.
4. **Write or restructure around the promise.** Use the type-specific guidance
below. Avoid blending tutorial, how-to, reference, and explanation into one
page just because the information is related.
5. **Verify the result.** Check that each section helps the primary reader job.
Remove or relocate sections that serve a different job.
6. **Polish the prose.** Run this step only when the deliverable contains prose
you wrote or rewrote. Skip it for classification notes, outline reviews,
and rewrite plans that contain no rewritten text. Run it last: structure
first, prose second.
- If a skill named `humanizer` is available in this session, invoke it on
the finished document and ask it to return only the final text, which
humanizer calls embedded mode. Tell it the Diataxis type so it keeps
that type's voice: guiding for a tutorial, imperative for a how-to,
neutral for reference, and weighing alternatives for an explanation.
Change prose only. Keep headings, section order, code, commands, tables,
and link targets unchanged.
- If it is not available, apply the Reader-First Prose rules and the
checklist in `references/ai-prose-tells.md`. Then tell the user once per
conversation that they can install the full skill with
`npx skills add blader/humanizer`. Do not repeat the suggestion.
- If Simplified Technical English applies, check the text against
`references/ste-rules.md` after the humanizer pass. STE is the last pass.
## Reader-First Prose
Write for comprehension rather than formality or literary flow.
- Express one main idea per sentence.
- Prefer 8–20 words per prose sentence.
- Review every sentence longer than 25 words. Split it unless doing so would
weaken accuracy or obscure the relationship between ideas.
- Prefer simple sentences. Use a subordinate clause only when it communicates
a necessary condition, cause, contrast, or qualification more clearly than
two separate sentences.
- Avoid nested subordinate clauses and sentences containing more than two
clauses.
- Move secondary details into a new sentence, list, note, or linked document.
- Preserve technical terms, contracts, constraints, evidence, and important
exceptions. Concision must not remove required meaning.
- Keep the certainty of each claim. Do not turn "may fail" into "fails". Do
not add a cause, frequency, or fact that the source does not state.
- Use one term for one concept. Do not rotate synonyms such as "check",
"verify", and "confirm" for the same action.
- Keep noun clusters to three words or fewer.
- Give each paragraph one topic and at most six sentences. Use a list for
three or more steps or conditions.
- Do not apply sentence-length limits to code, commands, paths, tables,
headings, quotations, or generated identifiers.
- Vary sentence length naturally. Do not turn every sentence into the same
short pattern. Text under Simplified Technical English is the exception.
Treat word counts as review signals, not mechanical pass/fail rules. For
Vietnamese, count space-separated tokens consistently even though they do not
always correspond to lexical words. For languages without space-delimited
words, rely on the one-idea and clause-count rules instead.
## Simplified Technical English
ASD-STE100 Simplified Technical English (STE) is a controlled language forFrom tronghieu/agent-skills, MIT.
More from tronghieu/agent-skills
- 0.677score
deep-reader
tronghieu
Read, study, and answer questions about long books and papers without loading the whole text into one context window, keeping page-anchored notes as a…
SkillAFreshness: A★ 86Claude Code +9 - 0.677score
manual-tester
tronghieu
Plan and run manual tests of the project's running app.
SkillAFreshness: A★ 86Claude Code +9 - 0.677score
slidewright
tronghieu
Build interactive presentation websites — slide decks projected to a room, controlled by one presenter.
SkillAFreshness: A★ 86Claude Code +9 - 0.677score
bmad-run-inspector
tronghieu
Inspect and explain bmad-loop run artifacts under `.bmad-loop/runs/`.
SkillAFreshness: A★ 86Claude Code +9 - 0.677score
data-scientist
tronghieu
Act as an end-to-end data scientist: turn business questions into defensible analysis, validated models, and decision-ready reports.
SkillAFreshness: A★ 86Claude Code +9 - 0.637score
scrum-master
tronghieu
Act as the team's Scrum Master of record, or copilot to a human one: plan sprints, run daily pulses, close sprints, prepare and follow up retrospectives…
SkillBFreshness: B★ 86Claude Code +9 - 0.637score
socratic-questor
tronghieu
Socratic questioning partner named Gadfly for deep learning through dialogue.
SkillBFreshness: B★ 86Claude Code +9 - 0.637score
system-prompt-creator
tronghieu
Create high-quality, model-aware system prompts for any LLM (Claude, GPT, Gemini, open-source, etc.).
SkillBFreshness: B★ 86Claude Code +9 - 0.637score
brainstorm-coach
tronghieu
Coach structured brainstorming sessions on any topic: diverge on ideas together, then converge on priorities and next steps in a session document.
SkillBFreshness: B★ 86Claude Code +9 - 0.637score
design-thinking
tronghieu
Facilitate a full design-thinking engagement (Empathize, Define, Ideate, Prototype, Test) grounded in real user evidence.
SkillBFreshness: B★ 86Claude Code +9 - 0.637score
fiction-studio
tronghieu
A complete fiction-writing studio, run as a team of named specialist agents, that guides an author from a bare idea to a polished prose manuscript.
SkillBFreshness: B★ 86Claude Code +9 - 0.637score
market-researcher
tronghieu
Conduct rigorous, citation-backed desk market research: quick go/no-go market scans, market sizing (TAM/SAM/SOM), competitor analysis, customer demand…
SkillBFreshness: B★ 86Claude Code +9
Signals
- Stars
- 86
- Forks
- 32
- Open issues
- 0
- Last commit (this folder)
- Oct 4, 2026 (25 days ago)
- Last release
- Sep 19, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/tronghieu/agent-skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from tronghieu
- 0.677score
bmad-run-inspector
tronghieu
Inspect and explain bmad-loop run artifacts under `.bmad-loop/runs/`.
SkillAFreshness: A★ 86Claude Code +9 - 0.677score
slidewright
tronghieu
Build interactive presentation websites — slide decks projected to a room, controlled by one presenter.
SkillAFreshness: A★ 86Claude Code +9 - 0.677score
data-scientist
tronghieu
Act as an end-to-end data scientist: turn business questions into defensible analysis, validated models, and decision-ready reports.
SkillAFreshness: A★ 86Claude Code +9 - 0.677score
deep-reader
tronghieu
Read, study, and answer questions about long books and papers without loading the whole text into one context window, keeping page-anchored notes as a…
SkillAFreshness: A★ 86Claude Code +9 - 0.677score
manual-tester
tronghieu
Plan and run manual tests of the project's running app.
SkillAFreshness: A★ 86Claude Code +9 - 0.637score
cv-scorer
tronghieu
Score candidate CVs on a 100-point scale against a Job Description.
SkillBFreshness: B★ 86Claude Code +9
Cite this page
diataxis-writer skill. Agent Mart. https://agentmart.dev/skills/tronghieu-diataxis-writer (last updated 2026-10-24)