Docsbook Skills

Catch a11y issues in your docs before users hit them. Audits markdown for WCAG 2.1 AA violations a screen reader actually cares about — missing alt text, broken heading hierarchy, vague link anchors, captions and code-block language. Returns a prioritized fix list per page.

Run a full audit of your documentation in one command. Orchestrates 10 specialized analyses — content type, structure, style, audience, links, SEO, a11y, i18n, media, maintenance — and returns a single prioritized report you can hand to the writer. Works on one page, a folder, or the entire docs/ tree.

Find out where your docs talk past their reader. Detects vocabulary mismatch, undeclared prerequisites, mixed-audience pages and jargon density against the stated reader profile — so junior devs stop bouncing on senior-level pages.

Stop mixing tutorials with reference. Classifies each page against the Diátaxis framework (tutorial / how-to / reference / explanation), flags misclassified or hybrid pages, and reports structural violations — the #1 reason users can't find answers in your docs.

Stop letting translations rot silently. Audits multilingual docs for content parity across languages, ISO 639-1 correctness, hreflang tags, navigation translation, date/number localization and out-of-date pages — across the 15 languages Docsbook supports. Skipped automatically if only one language is enabled.

Surface the docs that are quietly lying to your users. A quarterly-style audit that flags stale content, deprecated pages without migration paths, TODO/FIXME left in published docs, expired promises, old version references and ownership gaps. Not a single-page review — designed for the whole tree.

Find the bloated images and stale screenshots dragging your docs down. Scans every image, video and diagram referenced from markdown — bad formats, oversized files, missing alt text, screenshots older than the current UI — and returns a fix list per file.

Fix the dead ends and orphan pages in your docs. Walks the full doc graph to find broken internal links, pages nothing links to, generic anchor text ("click here"), over-deep hierarchies and missing next-step links between related pages. Cross-file — requires the whole graph, not one page.

Make your docs rankable — by Google and by AI. A documentation-aware SEO audit covering title and description optimization, heading hierarchy, keyword placement, internal topic clusters, image alt text and GEO / AI Overviews compatibility. Not a general site audit — tuned for doc-specific failure modes.

Catch malformed pages before reviewers do. Checks each page for frontmatter completeness, heading hierarchy, prerequisites section, code block language tags and length — and reports issues as machine-readable JSON without editing files. Cheap, deterministic, runs on every page.

Tighten the prose of your docs without rewriting by hand. Flags passive voice, filler words, marketing adjectives, runaway sentences, missing second person and inconsistent terminology — and returns issues per page without touching the files. Pair with /docs-create or /docs-analyze.

Set a Docsbook workspace's brand — accent/muted colors (light + dark), Google Font, theme, logo, icon, name — by deriving values from real signals instead of inventing them. Pulls the source site's theme-color and og:image, the dominant color of the logo or icon, branding already on the workspace, and brand mentions in the README, then proposes contrast-checked palettes and asks the user to choose. Never writes a value the user did not confirm. Pair with /docs-create, /docs-setup-workspace, /docs-style-tone, or /docs-analyze.

Turn a URL or repo into a live docs site in one command. Full end-to-end pipeline — detects the source (website, code, or Mintlify/GitBook/Docusaurus), generates structured Markdown, publishes to GitHub, and configures the Docsbook workspace. Minimal questions, maximum output.

Same as /docs-create, but with you in the loop. Pauses at six checkpoints — source detection, structure review, content enrichment, branding, languages, domain — so you can adjust before publishing. Use when you want full control over what ships.

Identify what kind of docs source you have in one call. Detects website URLs, code repositories, and existing platforms (Mintlify, GitBook, Docusaurus, Nextra) so the right builder is picked. Used internally by /docs-create or directly when you're not sure where to start.

Make a freshly-generated docs site rich and on-brand instead of a bare skeleton. Bundles two passes in sequence — (1) auto-brand from the source project's real signals (README, website theme-color, og:image, logo/favicon, repo name), and (2) enriched-structure generation that produces a multi-section site (getting-started, concepts, guides, reference, hero/landing) from audience, style-tone, content-type, and navigation heuristics. Run during first-generation, before publishing. Pair with docs-create.

Build Markdown docs from a GitHub URL or code repository — README, source tree, exported APIs, examples, comments. Use when the user provides a GitHub URL, repo URL, or says "from GitHub", "from repo", "import repo". Produces docs-output/<name>/ then publishes to GitHub and configures the Docsbook workspace automatically.

Import existing documentation from Mintlify, GitBook, Docusaurus, Nextra, or VitePress into a clean docs-output/<name>/ folder ready for Docsbook. Preserves structure and content; normalises platform-specific syntax to plain Markdown. Use when migrating away from another docs platform.

Spin up Markdown docs from any product website URL or live site. Crawls the site, extracts content, and produces a clean docs-output/<name>/ folder — then publishes to GitHub and configures the Docsbook workspace automatically. Use when the user provides a website URL, site URL, or pastes a link and says "from website", "import site", "from URL", "from a live site", "recreate this site's docs", "copy this link", "повтори документацию по ссылке", or "сделай доки по этому сайту".

Generate a complete, conversion-optimised docs site from a product name or idea alone — no URL, no repo needed. Invents the right pages, messaging, and structure for your product, then publishes and configures the Docsbook workspace automatically. Use when the user says imagine docs, create docs for X, придумай документацию, invent docs, make up docs, from scratch, no source, for better selling, marketing-grade docs, wow me, or provides only a product name with no URL.

Give every AI agent in your repo the same context about your docs. Reads your Docsbook workspace and writes an AGENTS.md at repo root so Claude Code, Cursor, Codex, Copilot and Gemini CLI all start each session knowing the docs URL, languages, AI-chat conventions and available @docs-skills commands.

Push a local docs folder to GitHub in one step. Handles git init, commit, gh repo create and push — no Docsbook MCP required. Designed as the natural next step after /docs-from-site, /docs-from-code or /docs-from-docs.

Configure a fresh Docsbook workspace from one command. Wires branding, UI toggles, AI chat, SEO, languages and custom domain via Docsbook MCP — the natural follow-up to /docs-publish. Falls back to printed setup instructions if MCP isn't connected.

Enable AI auto-translation for a Docsbook workspace across up to 15 languages in one command. Validates inputs, checks the workspace plan, enables the requested languages, switches translation mode to auto, and optionally registers a Slack webhook that fires when each translation batch completes. Requires PRO plan.

Add a CI gate to every pull request that checks whether code changes are accompanied by documentation updates, validates frontmatter in changed markdown files, and detects broken internal links. Generates a ready-to-use GitHub Actions workflow file.

Wire up release announcements for a Docsbook workspace. Registers a Docsbook webhook on release events and generates a GitHub Actions workflow that dispatches notifications to Slack and/or email when a new release is published. Requires PRO plan.

Register a Docsbook content.outdated webhook and generate a GitHub Actions workflow that converts each stale-content notification into a GitHub Issue in the documentation repository. Keeps docs maintenance visible in the team's normal issue triage flow. Requires PRO+ plan.

Detect and fix code↔docs drift before every push. Orchestrates a four-subagent pipeline — Haiku planner clusters changed code files, Haiku searchers find drifted pages per cluster in parallel worktrees, Sonnet editors rewrite the affected sections, and a Sonnet curator merges everything atomically into the commit. Works as a manual command or as an auto-installed pre-push git hook.

Bypass Docsbook's built-in AI translator and delegate translation work to a custom external service over webhooks. Switches the workspace to external translation mode, registers a translation.requested webhook, and scaffolds a handler function the user deploys with their own translation logic. Requires PRO+ plan.

Improve the AI chat system prompt of a Docsbook workspace using real negative feedback and unanswered questions from the last 30 days. Clusters failure patterns by topic, proposes a minimally invasive prompt update, shows a before/after diff, and applies the change only after explicit user confirmation. Requires PRO plan.

Splits high-dwell-time pages into "engagement signal" (deep interest) vs "engagement problem" (stuck users) by cross-referencing dwell time with negative feedback. Identifies which long-read pages are loved vs which are confusing. Produces an insight JSON report consumable by downstream actor agents. Requires PRO+ plan.

Maps the most common 3-step navigation journeys through your docs and identifies high-volume paths with low completion rates (users dropping before reaching a conversion page). Surfaces "broken journeys" — transitions implied by the doc graph that users do not actually take. Produces an insight JSON report consumable by downstream actor agents. Requires PRO+ plan.

Know exactly which docs page to write next. Cross-references real user signals — failed searches, AI-unanswered questions, popular queries — against your live doc graph and returns the top 7 pages worth creating. Optionally opens a GitHub Issue with a draft outline for each. PRO+.

Measures click-through rate on every internal link and CTA button across the docs. Flags conversion-critical buttons (Upgrade, Sign up, Book demo) whose CTR sits below the site median given comparable impressions. Also surfaces "orphan traffic" — pages with pageviews but zero outgoing clicks. Produces an insight JSON report consumable by downstream actor agents. Requires PRO+ plan.

Clusters every user question asked to the AI chat into themed topic groups, then cross-references against the doc graph to label each cluster as content_gap (no doc covers it) or ai_chat_failure (doc exists but chat couldn't surface it). Goes deeper than docs-gap-finder by including answered-but-not-helpful questions. Requires PRO plan.

Maps UTM-tagged traffic against landing pages to find mismatches between marketing promise and documentation reality. Surfaces UTM campaigns where users land but bounce — the doc page doesn't answer what the ad/post promised. Produces an insight JSON report consumable by downstream actor agents. Requires PRO+ plan.

Drills into the top-N most active anonymous visitors and clusters them by behavior pattern — "buyer-blocker" (visits pricing, leaves negative feedback), "tire-kicker" (browses but never reaches CTA), "deep-reader" (long engagement, positive signals). Surfaces cohort-level findings that page-level analytics miss. Produces an insight JSON report consumable by downstream actor agents. Requires PRO+ plan.