docs
Plan and write user documentation — Diátaxis structure, audience first, a reader test before merge.
Skill by joewalnes
Chalk Labs
Need help deploying docs?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
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/joewalnes-docs
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: testing, 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. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/docs
What's inside
- Skill name
docs- Description
- Plan and write user documentation — Diátaxis structure, audience first, a reader test before merge. Use when creating or restructuring docs.
- License
- MIT
- Instructions
- 52 lines in SKILL.md · view on GitHub
Files (2)
- SKILL.md
- writing-standard.md · references/
SKILL.md (first lines)
# Docs: write for the reader who arrived from a search engine
Documentation fails in one specific way when agents write it: it comes out as development notes. Pages open by describing how they were generated and verified, narrate the project's history, stack three facts into one sentence, and mix a tutorial's hand-holding with a reference's precision until they do neither. Every one of those pages was technically accurate. A human read one and said "the text is gibberish — how are they meant to understand it?" Four pages were rewritten. This skill exists so that happens zero times, not once.
## Before writing anything: research, then plan
If this project has never had real docs, do not start by writing pages. **Look at how three comparable projects document themselves** — structure, tooling, what a first-time reader lands on — and write down what you'd take from each. Then plan:
1. **Who reads this, in the order they show up.** Name them: the evaluator deciding in ten minutes, the newcomer with a terminal open, the implementer who needs the exact contract, the operator deploying it, the stuck person who needs the answer in thirty seconds. Every page is written for one of them.
2. **Inventory what needs documenting** from the code, the issue tracker's recurring questions, and the old docs — recurring support questions are the highest-value pages.
3. **Choose the four Diátaxis sections** (below) and place every inventory item in exactly one.
4. **Tooling as an options table** — generator, hosting, single-sourcing of generated reference (flags, config, API) so it *cannot* drift, link checking — with a recommendation, and **the actions only the human can take** (a DNS record, a domain, a secret) called out first, so they're never a surprise.
The plan is a file in the repo. Get it read before building.
## The structure: Diátaxis, and why a page is only one thing
| Mode | Reader's state | The page answers | Shape |
|---|---|---|---|
| **Tutorial** | "I've never used this" | "Take me through it once, successfully" | Numbered steps, one path, guaranteed outcome, no options |
| **How-to** | "I have a specific job" | "The steps for this exact task" | Task-titled, fix first, theory linked not included |
| **Reference** | "I need a precise fact" | "What exactly does X do?" | Complete, consistent, generated where possible, no advice |
| **Explanation** | "I want to understand why" | "Why does it work this way?" | Prose, context, trade-offs, no steps |
A page that serves two modes serves neither. The commonest failure is a how-to that teaches, or an explanation that drifts into steps. When a page starts doing both, split it and link — each side links the other, neither repeats it.
## The writing standard
The binding rules are in `references/writing-standard.md`; read it before writing a page. The five that catch the most defects:
- **Never write about the documentation.** No page or section opens by describing how it was produced, generated, or verified. Provenance that helps ("generated — don't hand-edit") is one line at the foot.
- **Front-load the answer.** A reader who stops after one sentence still has it.
- **One idea per sentence.** Three technical facts in one sentence is the house defect.
- **Conditions before instructions.** "If you're behind a proxy, set X" — not the reverse.
- **Define jargon on first use, or link it.** They did not arrive from the previous page.
## The reader test — the gate
Before any docs page merges, **someone who has not seen the code reads it as the target reader** and answers two questions: *what is this page for?* and *what would I do next?* In a fleet, that's a fresh agent with no context and the persona from the plan. If it can't answer, the page goes back — not to be verified against the code again (it was already accurate), but to be rewritten for a person. Run it on your own drafts too: read the page aloud as the newcomer.
## Mechanisms, not intentions
- **Generated reference has a drift check that fails.** If flags, config or API are documented, generate that page from the source of truth and add a CI check that fails when the generated output and the committed page differ — and prove it fails on a hand edit before trusting it.
- **Links are checked in CI**, and the check is proven to fail on an injected broken link.
- **`/llms.txt`** (and `/llms-full.txt`) alongside the site, so agents reading the docs get the plain-text form.
- The writing standard lives in the repo (`docsite/STYLE.md` or equivalent) and says: *if a page breaks a rule here, the page is wrong.*
## Reviewing existing docs
`review <page-or-dir>`: for each page, name its mode (or "two modes — split"), its reader, whether the first sentence is the answer, every sentence that talks about the documentation itself, every undefined term, and whether a stuck reader would find it in thirty seconds. Then apply the reader test. Report as a table, worst first.From joewalnes/skills, MIT.
More from joewalnes/skills
- 0.568score
request
joewalnes
Ask another project's agents for a change instead of editing their repo.
SkillBFreshness: B★ 3Claude Code +9 - 0.566score
delegate-image
joewalnes
Generate a visual — SVG via text models (Astra) for logos and diagrams, a raster panel for photos — judged by two independent models.
SkillBFreshness: B★ 3Claude Code +9 - 0.566score
slop
joewalnes
Find unchosen code — surface area growing without leverage.
SkillBFreshness: B★ 3Claude Code +9 - 0.564score
delegate-review
joewalnes
Independent code review of Claude's own work by a different model lineage (Kimi) via pi.
SkillBFreshness: B★ 3Claude Code +9 - 0.564score
fleets
joewalnes
Which go-team fleets are alive, across every project — leases, last landing, session state.
SkillBFreshness: B★ 3Claude Code +9 - 0.563score
project-setup
joewalnes
Walk through project setup improvements for efficient AI-assisted development
SkillBFreshness: B★ 3Claude Code +9 - 0.563score
sitrep
joewalnes
Where were we?
SkillBFreshness: B★ 3Claude Code +9 - 0.558score
delegate-private
joewalnes
Run a task on a local Ollama model so data never leaves the device.
SkillBFreshness: B★ 3Claude Code +9 - 0.558score
bug-bash
joewalnes
Work through a project's bug list autonomously, fixing bugs in priority order
SkillBFreshness: B★ 3Claude Code +9 - 0.525score
bug
joewalnes
Alias for /todo — add a bug to the tracker.
SkillBFreshness: B★ 3Claude Code +9 - 0.443score
release-setup
joewalnes
Set up automated cross-platform binary releases for a Go project
SkillDFreshness: D★ 3Claude Code +9 - 0.443score
import-skill
joewalnes
Discover and import Claude Code skills into this shared repo
SkillDFreshness: D★ 3Claude Code +9
Signals
- Stars
- 3
- Forks
- 3
- Open issues
- 0
- Last commit (this folder)
- Sep 7, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/joewalnes/skills
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from joewalnes
- 0.568score
request
joewalnes
Ask another project's agents for a change instead of editing their repo.
SkillBFreshness: B★ 3Claude Code +9 - 0.566score
delegate-image
joewalnes
Generate a visual — SVG via text models (Astra) for logos and diagrams, a raster panel for photos — judged by two independent models.
SkillBFreshness: B★ 3Claude Code +9 - 0.566score
slop
joewalnes
Find unchosen code — surface area growing without leverage.
SkillBFreshness: B★ 3Claude Code +9 - 0.564score
fleets
joewalnes
Which go-team fleets are alive, across every project — leases, last landing, session state.
SkillBFreshness: B★ 3Claude Code +9 - 0.564score
delegate-review
joewalnes
Independent code review of Claude's own work by a different model lineage (Kimi) via pi.
SkillBFreshness: B★ 3Claude Code +9 - 0.563score
project-setup
joewalnes
Walk through project setup improvements for efficient AI-assisted development
SkillBFreshness: B★ 3Claude Code +9
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get docs running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
docs skill. Agent Mart. https://agentmart.dev/skills/joewalnes-docs (last updated 2026-10-24)