aidou

Build and publish the book from one Markdown source to a GitHub Pages website, a print-quality PDF, and a reflowable ePub, using Quarto. Covers where content/art/tokens live (book/, images/, theme.yml), regenerating the Quarto pages, local render/preview, Mermaid + GFM-alert + numbering handling, cover and chapter-art generation, and the GitHub Actions deploy. Use when rendering, previewing, exporting, deploying, or restyling the book.

Bump the book's edition ("Edition X.Y") and add a Changelog entry, keeping the version string in sync across the web landing page, the PDF copyright page, the ePub colophon, and the Changelog. Use whenever the user wants to release/cut a new edition, bump the version, or update the changelog — e.g. "release Edition 1.1", "bump the version", "cut a new edition", "upgrade the version".

Apply and audit APA 7 citations across a Markdown book. Rewrites in-text citations to hyperlinked author–year (author + title on first mention, plain on later mentions), regenerates a per-chapter References section, and audits for internal-download links, orphans, and broken link text. Use when adding or reformatting citations, adding a source, or building the reference lists.

Given a plain-language description, find the best-matching open-source illustration on undraw.co, recolour it to the book's Rosely palette, save it, and (if a section is named) place it in the book. Use whenever the user wants to "find an image/illustration/graphic for" a spot, "add a picture to §X", "get an undraw image of …", or otherwise illustrate a passage. Because an exact match is unlikely, the skill is deliberately resilient: it tries several search terms, downloads a shortlist of ~3 candidates, renders a preview of each, looks at them, and picks the one that actually matches the description — not just the closest slug. undraw illustrations are flat vector SVGs released free for commercial use with no attribution, and every accent is one recolourable colour, so the winner can be restyled to the book's per-chapter accent automatically. The chain for one image: description → several search terms → pooled candidates → download 3 SVGs → render PNG previews → evaluate and select the best → recolour the win

Ingest a single cutting-edge article, post, or talk — an X (Twitter) link, a blog post, a newsletter, a LinkedIn piece, or a YouTube video (transcript) — and weave it into the book as one well-placed, honestly-flagged citation. Use whenever the user hands you a link (especially an opinion piece, a practitioner thread, an industry-leader post, or a conference talk) and says "treat this as a source", "ingest this", "add this article", or "cite this". These sources are usually emerging opinion rather than verified research, so the citation must say so. The skill runs the whole chain for one link: fetch (web pages via download-source; X/Twitter via the fxtwitter JSON mirror; YouTube via the youtube-transcript-api tool) → save to sources/ → summarise into summaries/ → find the single best home in book/*.md → cite it (flagged as emerging opinion) → update references.json → run the book-style loop on just that section → verify and build. It is the lightweight, single-link cousin of research-topic, and it reuses down

Number Markdown section and subsection headings as N.M and N.M.K per chapter (e.g. 1.2, 2.3.1), leaving chapter titles and the References section unnumbered and skipping code fences. Idempotent — safe to re-run after adding, removing, or reordering sections. Use when adding numbered headings to a book or renumbering after edits.

Rewrite a section or chapter into the house prose style, then loop a self-check until it stops sounding like AI. Use whenever the user wants a style pass, a de-AI pass, a rewrite "in my voice", or any drafting or revising of chapter prose — or names a chapter/section and asks to fix its tone, readability, or style. The style is modelled on Kernighan & Pike: plain, explanatory, one professional walking another through an idea, with a slight university-professor tone. The loop rewrites, reads the result against the style guide asking "does this sound like AI?", fixes what it finds, and checks again until a pass comes back clean. Also carries the style rules and an advisory prose linter.

Weave a source into the book — cite it and put its evidence to work in the chapters. Use whenever the user wants to add/integrate/cite a source into the book, or after summarise-source has written a summary. Given a summarised source (a file in summaries/, or a source the user names), it finds where the source belongs in book/*.md, adds an inline APA citation plus a References entry, and updates references.json. If the user names a section it focuses there; otherwise it finds the right place(s), which may be several. If the source is *already* cited, it finds every place the book uses it and reviews each citation against the (possibly newly regenerated) summary, tightening claims, numbers, placement, or citation form where the source now supports something better. If the source isn't relevant, or the point is already well covered by another source, it says so instead of forcing a cite.

End-to-end research pipeline for the book: take a topic all the way from discovery to integrated-and-cited. Use whenever the user wants to "research and add a topic", "find and integrate sources on X", "do a research pass on Y", or otherwise work a whole subject into the book. It orchestrates four sub-skills in order — search-topic (find sources) → download-source (fetch into sources/) → summarise-source (write each summary) → enhance-book (cite them into the chapters). Give it a topic, or just a section to enhance — with no topic it reads the section and derives the topics to research. It reuses sources already in the repo (refreshing any stale or dossier-style summary) as well as fetching new ones. Each sub-skill is also usable on its own; reach for this one when the user wants the full chain, not a single stage.

Review a whole book chapter over several passes to get it publication-ready. Use whenever the user wants to review, polish, sanitise, finalise, proof, tighten, or "make publication-ready" a chapter or the preface — or names a chapter/section and asks to "clean it up", "do a pass on it", "get it ready to publish", or "check it over". It runs seven checks: grounded citations (every claim and quote traces to a cited primary source), definitions (key terms defined on first use — a GFM alert for the important ones, inline for the peripheral), thin sections (too few citations → research and thicken via research-topic), tables (prose that would scan better as a table), diagrams (relationships/processes worth a Mermaid figure), cross-chapter consistency (terms, citation labels, cross-references and claims kept in step with the rest of the book), and a final readability, consistency, Australian-spelling and flow pass. It audits first and shows you the findings for a go-ahead before applying fixes, because the fix stag

Turn a downloaded source document into a dense, faithful study-guide summary in the repo's summaries/ folder. Use whenever the user wants to summarise, digest, condense, or write a Cliff's-Notes / study-guide version of a source in sources/ (a Markdown file, an arXiv or other PDF, or any other downloaded document) — or wants to (re)generate, backfill, or refresh a file in summaries/. Trigger even if the user only names a file, a paper, or a topic under sources/ and asks to "summarise it", "write it up", "make a summary", or "create a dossier".

Download a source into the repo's sources/ tree, in the right form and place. Use whenever the user wants to fetch/save/grab a source — an arXiv paper, a web page or article, an analyst PDF — given a source from search-topic or a URL/arXiv id the user names. Puts arXiv papers as PDFs in sources/arXiv/, converts HTML pages to Markdown in the right sources/<topic>/ folder (creating the topic folder if new), and downloads other PDFs into sources/<topic>/. Reports the local path. Pair with search-topic (before) and summarise-source (after).

Do deep research on a topic and return a curated list of sources relevant to the book. Use whenever the user wants to find or research sources on a subject — "find papers on X", "what's the research on Y", "search for sources about Z", "any good sources on ...". Prioritises arXiv and other academic/primary sources but also returns Wikipedia (definitions/history), analyst and industry reports, and reputable blog/news articles. Returns a ranked list (topic, title, type, year, URL, why-relevant) — it does NOT download; pair with download-source next.

Create or edit C4 architecture diagrams in Mermaid. Use for system context, container, component, and code-level architecture views, updating existing .mmd files, revising Mermaid blocks in Markdown, or turning architecture notes into C4 diagrams.

Use when creating, editing, or generating draw.io diagram files (.drawio, .drawio.svg, .drawio.png). Covers mxGraph XML authoring, shape libraries, style strings, flowcharts, system architecture, sequence diagrams, ER diagrams, UML class diagrams, network topology, layout strategy, the hediet.vscode-drawio VS Code extension, and the full agent workflow from request to a ready-to-open file.

Create or edit draw.io diagrams. Use for architecture maps, process flows, operating models, decision trees, and strategy visuals in .drawio files, then export to PNG or SVG for sharing.

Comprehensive guide for creating software diagrams using Mermaid syntax. Use when users need to create, visualize, or document software through diagrams including class diagrams (domain modeling, object-oriented design), sequence diagrams (application flows, API interactions, code execution), flowcharts (processes, algorithms, user journeys), entity relationship diagrams (database schemas), C4 architecture diagrams (system context, containers, components), state diagrams, git graphs, pie charts, gantt charts, or any other diagram type. Triggers include requests to "diagram", "visualize", "model", "map out", "show the flow", or when explaining system architecture, database design, code structure, or user/application flows.

Convert Mermaid diagrams into draw.io XML with high structural fidelity. Use when you need editable .drawio source from Mermaid code by analyzing nodes, edges, groups, and labels, then generating accurate draw.io output.

Write and execute Python scripts in this workspace, installing any packages they need with conda or pip. Use for data analysis, charts, quick computations, scraping/parsing research sources, validating numbers cited in the book, or any one-off automation that benefits from running real Python.