---
name: write-docs
description: Write or revise a documentation page for its content type, such as a tutorial, how-to guide, reference, or explanation. Use when drafting, restructuring, or reviewing a docs page, or when it's unclear whether the reader or the system should be the actor.
license: MIT
metadata:
  author: influxdata
---

Load the gdd-style skill first.
Its rules apply to every page.
This skill adds the rules that depend on the content type.

## 1. Name the content type

Decide the page's content type before you write or revise it.
If the page mixes types, tell the writer and suggest a split.

| Content type | The reader wants to | Who acts | Voice |
| --- | --- | --- | --- |
| Tutorial | learn by doing | the reader | Imperative steps and "you." One path, no alternatives. |
| How-to guide | finish a task they already understand | the reader | Imperative steps. State each condition before its step. |
| Reference | look up a fact | the system: a server, API, command, or component | Name the component. Passive voice is fine when the actor is unknown or irrelevant. Use "you" only for what the reader supplies or controls. |
| Explanation | understand why something works as it does | the system or the design | Name the component or the decision. Keep reasons and tradeoffs. |

## 2. Make the actor clear

- **When the reader acts, say so.**
  Use the imperative or "you."
  Domain experts often write "the column is added" when the reader adds it.
  In a tutorial or how-to guide, rewrite that as "Add the column."
- **When the system acts, name the part that acts:**
  "The server rejects the write," not "The write is rejected."
- **When you can't tell who acts, keep the passive.**
  Flag the sentence for the writer.
  A technical writer or the page's owner makes that call.

When you revise, gdd-style's fidelity check still applies.
Making the reader the actor of a step the source already describes isn't a new claim.
Naming a system actor the source doesn't state is.

## 3. Check before you finish

The page is done when:

- it has one content type, or you've flagged the mix;
- every step and every system behavior has an actor, or a flag;
- the voice matches the row for its content type.