explainer

explainer is a Claude Code plugin by mizchi (Kotaro Chikuba) for explaining things to people who can't keep up with what a coding agent wrote. It builds a picture of one reader, writes only what that reader doesn't already know, and checks the quoted outputs and the figures with tools before you share the result. It's a community project, published from his GitHub repository. What its skills teach your agent - explainer: a crash course for one reader. It builds a persona of what they know from questions and public information, writes only the difference, re-runs every output it quotes, and draws figures in Mermaid, D2, SVG or HTML that are checked against a fact sheet and for overlaps, clipping and arrow readability, then looked at and fixed - explainer-book: a chaptered course, with learning objectives matched to quizzes, no concept used before it's introduced, a reading-time budget per chapter, and exercises whose answers are run - first-reader: simulated beta readers who read a draft one passage at a time and report where they drop off and what they remember the next day. It doesn't rewrite - d2-diagram: an architecture, data-model or call-flow diagram in D2, laid out by TALA, read in the terminal and held to a fact sheet - d2-slides: a slide deck from one Markdown file with a D2 figure per slide, built to HTML and checked for slides that break, clip or lack contrast How to install it In Claude Code, run /plugin marketplace add mizchi/explainer, then /plugin install explainer@explainer. The skills alone also install with npx skills add mizchi/explainer or with APM. Install explainer, explainer-book and first-reader together, because explainer-book runs explainer's checker. Good to know - The checking scripts need Node 24 or later, and npm i -D @mizchi/vlmkit marked playwright mermaid in the repository that holds your documents. Add the d2 CLI for D2 figures, and the Iconify lucide and logos sets for icons. first-reader needs only Python 3. - It also explains model or analysis results, such as diagnostics, metrics and charts, to the person who has to decide on them. - The explainer and explainer-book skills are written in Japanese. - first-reader is bundled from Shubhamsaboo/awesome-llm-apps and keeps its Apache-2.0 license; d2-diagram and d2-slides come from mizchi's vlmkit. The rest is MIT.

• “Explain this PR so our on-call SRE can understand what changes during an outage.”

  Copy

• “Write me a crash course on formal methods. I tried, but lost confidence as I wrote it.”

  Copy

• “Turn this into a chaptered course that starts with 01-quickstart.”

  Copy

• “Beta-read this draft: where would readers stop, and what would they remember tomorrow?”

  Copy

• “Draw our architecture as a D2 diagram we can keep in the repo.”

  Copy

• “Make a slide deck for this design review from one Markdown file.”