The /dr-write Command — Structured Content Creation
How /dr-write activates the Writer agent to plan, draft, and hand off content — from blog posts to technical docs — with a built-in quality pipeline.
Writing is easy to start and easy to do poorly. The /dr-write command exists to close that gap: it activates the Writer agent, which treats a content task the same way Datarim treats any other task — as a pipeline with defined stages, not a free-form prompt.
The command accepts a topic, a file path, or a brief description of what needs to be produced. From there, the Writer agent takes over.
What the Writer agent does
Before drafting anything, the agent answers four questions: what type of content is this, who reads it, what register does it need (technical, conversational, formal), and how long should it be for the target platform. Only after those are clear does drafting begin.
The pipeline runs in sequence. First, a research and planning pass: relevant sources are gathered, an outline is created, and any claims that will need verification are flagged. Then the draft is written section by section. After the draft is complete, a self-review checks structure, flow, and naturalness — catching the problems that slip through when writing fast.
What it produces
The output is a draft with editorial notes attached. Sections that contain claims needing external verification are marked. Sections where the prose felt mechanical during self-review are flagged for the next stage. The draft is not presented as finished — it is presented as ready for the editorial pass.
That hand-off is explicit. After /dr-write completes, the recommended next step is /dr-edit, which runs fact-checking and removes AI writing patterns. For smaller pieces that only need a targeted check, /factcheck or /humanize are available as lighter alternatives.
Supported content types
The Writer agent covers a wide range: technical documentation (READMEs, API docs, migration guides), blog posts and articles, social media posts adapted per platform, research writing, and formal business documents. The register and structure adjust to the type — a troubleshooting guide reads differently from a thought-leadership post, and the agent is aware of that distinction.
For social media specifically, the agent accounts for platform length limits before writing. A Telegram post has a 4096 UTF-16 unit ceiling; a LinkedIn post has different norms. Content that exceeds limits gets flagged for splitting rather than silently truncated.
Where it fits in the pipeline
/dr-write is the entry point of the content track in Datarim. Once a draft is approved through editorial review, /dr-publish handles formatting for each target platform and the mechanics of getting the content live.
Read what Datarim is for context on the broader pipeline, or see the /dr-edit command for what happens after the draft is ready.