Study Writing Style

This skill turns a writer's blog into a detailed, example-rich style guide that another agent (or the user) can use to rewrite articles in that writer's voice. The output is a single markdown file at writing-styles/<author-slug>.md, dense with verbatim quotes from the source.

When to use this skill

Trigger this skill when the user asks to study, analyze, or capture how a specific author writes, and is pointing at a real corpus (a blog index, an archive page, a list of posts). The goal is always to produce a reference document that someone could hand to a model and say "rewrite my draft in this voice". This skill does NOT rewrite the user's own articles; that's a separate step done with the output of this skill in context.

If the user only gives you an author's name with no link, ask for the blog/archive URL before starting. You cannot do this well from training data alone — the whole point is to ground every observation in verbatim text the author wrote.

What you need from the user

Before starting, confirm you have:

1. Author name — exactly how they want it spelled in the guide.
2. Blog index URL — the page that lists their posts. Archive page, "/blog/all", "/posts", or similar.
3. Output location (optional) — defaults to writing-styles/<author-slug>.md in the current working directory's project root. <author-slug> is the author's handle or kebab-cased name (tkdodo, julia-evans, dan-luu).
4. Project writing rules (auto-detected) — read the nearest AGENTS.md, then a project CLAUDE.md if no AGENTS.md is on the path, then the user-global ~/.claude/CLAUDE.md, for any writing-style constraints (no em-dashes, no emojis, sentence length caps). You will call these out in the guide as substitution rules when the studied author violates them.