docs-skills

52 skills that teach your AI agent to run your documentation like a growth channel.

Install once. Then just ask, in plain language.

Install • Find your pain • Full catalog • How it works • CLI

npx skills add Docsbook-io/docs-skills --skill '*'

Find your pain

Scan the left column. Find yours. Run the command.

😫 "My docs are a mess and I don't know where to start"

| Your pain | Skill | What you get back | |---|---|---| | I don't know what's broken | docs-analyze | One prioritized report — 10 audits in a single pass | | Users can't find answers | docs-content-types | Every page classified against Diátaxis, misclassifications flagged | | Readers bounce off the page | docs-audience | Jargon, undeclared prerequisites, mixed-audience pages | | Docs read like a brochure | docs-style-tone | Passive voice, filler, marketing adjectives — line by line | | Broken links, orphan pages | docs-navigation-linking | Full doc-graph walk: dead ends, orphans, "click here" anchors | | PRs land malformed pages | docs-structure-templates | Frontmatter, heading nesting, untagged code blocks | | Nobody finds us on Google | docs-seo | Audit judged against the queries you actually rank for — not guessed from the page text | | We get impressions, not clicks | docs-rank-recovery | Pages stuck at position 5–20, ranked into a rewrite queue — the cheapest traffic you can buy | | Screen readers can't use it | docs-accessibility | WCAG 2.1 AA violations that actually matter in markdown | | Translations rotted silently | docs-i18n | Content parity, ISO codes, hreflang, nav coverage | | Screenshots show the old UI | docs-media | Stale screenshots, bloated images, missing alt text | | Docs quietly lie to users | docs-maintenance | Stale content, deprecated pages with no migration path, TODOs |

npx skills add Docsbook-io/docs-skills --skill docs-analyze

"Audit my docs and give me one prioritized report"

🚀 "I have no docs — or the wrong ones"

| Your pain | Skill | What you get back | |---|---|---| | I have a repo/URL and 0 docs | docs-create | Detect source → generate → publish → configure. One command | | Same, but I want to approve each step | docs-create-interactive | Six checkpoints you control before anything ships | | I don't know what my source even is | docs-detect-source | Website / repo / Mintlify / GitBook / Docusaurus — identified | | Only a marketing site exists | docs-from-site | Rich conversion-grade Markdown crawled from the live site | | Only code exists | docs-from-code | README, source tree, exported APIs, examples → Markdown | | Trapped in GitBook/Mintlify | docs-from-docs | Clean portable Markdown, structure intact | | Only a product name exists | docs-imagine | A full invented docs site — pages, messaging, structure | | New site looks like a skeleton | docs-first-run-enrichment | Auto-branding + real getting-started and landing pages | | Docs look nothing like my product | docs-branding | Colors, font, theme, logo derived from real signals | | Docs inform but never sell | docs-sales-conversion | Monetization model detected → pricing page, CTA ladder, no dead ends | | Pages are walls of links | docs-content-widgets | Card grids and accordions from invisible markers — source stays plain markdown | | ChatGPT/Perplexity never cite us | docs-ai-retrieval | Passage-level rewrite patterns that survive controlled studies |

npx skills add Docsbook-io/docs-skills --skill docs-create

"Turn this GitHub repo into a live docs site"

📊 "I'm guessing what to write next"

Real user signal instead of opinion.

| Your pain | Skill | What you get back | |---|---|---| | What do we fix this week? | docs-health-triage | Every health signal merged into one ranked queue, each item handed to the skill that fixes it | | Which page do I write next? | docs-gap-finder | Failed searches + unanswered AI questions → ranked page list | | Search found it, nobody clicked | docs-title-rewriter | Rewritten titles and first lines, verbatim and ready to paste — no page rewrites | | Do our docs sell or just support? | docs-buying-blockers | Chat conversations split by buying stage: who's evaluating, what blocks them, which competitors they name | | Readers leave with nothing | docs-dead-end-hunter | Ranked give-up pages, with the actual journeys behind them | | What are people even asking? | docs-question-clusterer | Chat questions clustered, labeled content-gap vs retrieval-miss | | Long dwell time — good or bad? | docs-engagement-analyzer | Deep interest vs stuck users, split by negative feedback | | Where do users drop off? | docs-funnel-mapper | Common 3-step journeys, high-volume paths with low completion | | Nobody clicks my Upgrade CTA | docs-link-click-analyzer | CTR per internal link, conversion-critical buttons that underperform | | Campaign traffic bounces | docs-utm-analyzer | Where the ad promise doesn't match the doc reality | | Who are my power users? | docs-visitor-cohort | Top visitors clustered — including "buyer-blockers" |

npx skills add Docsbook-io/docs-skills --skill docs-gap-finder

"What are users searching for but not finding?"

🔁 "I keep doing the same maintenance by hand"

Install once, runs forever.

| Your pain | Skill | What you get back | |---|---|---| | Docs drift behind the code | docs-sync | Pre-push hook: detects drift, rewrites the affected sections | | PRs merge with stale docs | docs-pr-check | GitHub Action gate on every PR | | Stale pages nobody notices | docs-stale-watcher | Every outdated page becomes a GitHub Issue automatically | | We're launching a new market | docs-enable-translation | Up to 15 languages, auto-mode, Slack ping per batch | | I post release notes by hand | docs-release-announce | Slack/email fires on release: published | | AI chat gets thumbs-down | docs-tune-ai-chat | Failures clustered → a prompt update you approve | | I want my own TMS | docs-translate-webhook | Route translation to DeepL or your own pipeline |

npx skills add Docsbook-io/docs-skills --skill docs-sync

"Stop my docs from drifting out of sync with the code"

📦 "Now ship it"

| Your pain | Skill | What you get back | |---|---|---| | Docs sit on my laptop | docs-publish | git init → commit → gh repo create → push. No account needed | | Workspace config is 20 clicks | docs-setup-workspace | Branding, UI, AI chat, SEO, languages, domain — one command | | Every agent re-learns my docs | docs-generate-agents-md | AGENTS.md at repo root, so Cursor/Claude Code start informed |

🧭 "I need a plan before I write anything"

| Your pain | Skill | What you get back | |---|---|---| | I don't know where to start | docs-strategy-plan | Guided interview → what to write, for whom, in what order | | I don't know who actually buys | docs-audience-enricher | Buyer segments, entry paths, competitors — into your source-of-truth | | Every new page needs the same fixes | docs-authoring-rules | The rulebook loaded before you write — page type, structure, style, next step — so the audits find nothing |

Install (30 seconds)

Works in Claude Code, Cursor, Codex, Copilot.

npx skills add Docsbook-io/docs-skills --skill '*'

One skill only:

npx skills add Docsbook-io/docs-skills --skill docs-analyze

Then ask in plain language. No flags, no config — the skill fires from your request:

"Audit my docs and give me one prioritized report" "Turn this GitHub repo into a live docs site" "Which pages are users searching for but not finding?" "Translate my docs into 15 languages"

Before / after

| Without docs-skills | With docs-skills | |---|---| | Agent hallucinates doc structure | Agent works from your real docs — files, sitemap, or doc graph | | One-off prompts, rewritten per project | Reusable catalog that works everywhere | | "Looks good to me" | Every skill ships a pass/fail acceptance checklist | | Hours of prompting per audit | docs-analyze → one prioritized report | | Docs describe the product | Docs convert readers into users |

Full catalog

52 skills, 7 categories. Browse live: docsbook.io/skills.

| Skill | Plan | Ask it | |---|---|---| | docs-analyze | free | "Run a full health check before our release" | | docs-content-types | free | "Check my docs against Diátaxis" | | docs-structure-templates | free | "Validate the structure of every page in docs/" | | docs-style-tone | free | "Where am I using 'simply' and 'just' and 'powerful'?" | | docs-audience | free | "Which pages talk over the reader's head?" | | docs-navigation-linking | free | "Which pages are orphans with nothing linking to them?" | | docs-seo | free | "Why isn't this docs section getting organic traffic?" | | docs-title-rewriter | PRO | "Our search returns results and nobody clicks them — fix the titles" | | docs-accessibility | free | "Run an a11y audit before we launch publicly" | | docs-i18n | free | "Are my translations lagging behind the English source?" | | docs-media | free | "Which screenshots are out of date after the UI redesign?" | | docs-maintenance | free | "Find docs that reference features that no longer exist" |

| Skill | Plan | Ask it | |---|---|---| | docs-authoring-rules | free | "Load the rules before I write this page" | | docs-create | free | "Turn this GitHub repo into a live docs site" | | docs-create-interactive | free | "Build a docs site — pause before publishing so I can review" | | docs-detect-source | free | "Is this a website, a repo, or an existing docs platform?" | | docs-from-site | free | "Crawl this site and turn it into Markdown docs" | | docs-from-code | free | "Generate API docs from this codebase" | | docs-from-docs | free | "Migrate my docs off Mintlify" | | docs-imagine | free | "Imagine docs for my product — I only have a name" | | docs-first-run-enrichment | free | "Turn this skeleton into a real multi-section site" | | docs-branding | free · MCP | "Pull brand colors from my website and apply them" | | docs-sales-conversion | free | "Make my docs sell, not just inform" | | docs-content-widgets | free | "This page is a wall of links — make it a card grid" | | docs-ai-retrieval | free | "Rewrite this so ChatGPT and Perplexity can cite it" |

| Skill | Plan | Ask it | |---|---|---| | docs-gap-finder | free | "Which docs page should I write next?" | | docs-dead-end-hunter | PRO | "Which pages are losing people?" | | docs-question-clusterer | PRO | "Cluster my AI chat questions and tell me what's missing" | | docs-engagement-analyzer | PRO+ | "Which long-read pages are loved vs confusing?" | | docs-funnel-mapper | PRO+ | "Where are users dropping before a conversion page?" | | docs-link-click-analyzer | PRO+ | "Is my Upgrade button buried where nobody clicks it?" | | docs-utm-analyzer | PRO+ | "Which campaigns send traffic that bounces?" | | docs-visitor-cohort | PRO+ | "Who are my most active visitors and what are they doing?" | | docs-buying-blockers | PRO+ | "Who is evaluating us and what stops them from buying?" |

| Skill | Plan | Ask it | |---|---|---| | docs-sync | free | "Detect and fix code↔docs drift before this push" | | docs-pr-check | free | "Add a CI gate that flags code shipped without docs" | | docs-enable-translation | PRO · MCP | "Translate my docs into 15 languages" | | docs-release-announce | PRO · MCP | "Announce every release to Slack automatically" | | docs-tune-ai-chat | PRO · MCP | "Users keep thumbs-downing the AI chat — fix the prompt" | | docs-stale-watcher | PRO+ · MCP | "Turn stale-content alerts into GitHub Issues" | | docs-translate-webhook | PRO+ · MCP | "Route translation to my own DeepL pipeline" |

| Skill | Plan | Ask it | |---|---|---| | docs-publish | free | "Push my local docs folder to GitHub" | | docs-setup-workspace | free · MCP | "Configure my Docsbook workspace from one command" | | docs-generate-agents-md | free | "Generate an AGENTS.md so every AI agent knows about my docs" |

| Skill | Plan | Ask it | |---|---|---| | docs-strategy-plan | free | "I have no docs and don't know where to start" | | docs-health-triage | PRO | "What should we work on this week?" |

| Skill | Plan | Ask it | |---|---|---| | docs-rank-recovery | free | "Which pages are one rewrite away from real traffic?" | | docs-audience-enricher | PRO · MCP | "Figure out who actually buys my product and how they enter" |

Plan / MCP columns. Most skills are free and run on any agent with no account. MCP marks the few that genuinely operate a hosted workspace (branding, languages, webhooks) — they fall back to printed instructions when it's absent. PRO / PRO+ marks skills that read production analytics.

How it works

A skill is a reusable regulation — Workflow + Guardrails + Acceptance Criteria — not a one-off prompt.

Each skill states a need ("get the list of pages in scope and read their content") and how to reason about it, rather than a hardcoded way to fetch it. The agent satisfies the need with whatever it has connected:

AI Tool (Claude Code / Cursor / Codex / Copilot)
        │
        ▼
  docs-skill runs  ──▶  states the need + how to think
        │
        ├─ nothing connected ──▶  grep / find over the docs folder
        ├─ markdown-lsp       ──▶  semantic / graph search, self-hosted
        └─ Docsbook MCP       ──▶  same capability in the cloud + workspace settings
        │
        ▼
  Audit report / generated files / configured workspace

The agent sees what's available and picks the cheapest path.

The catalog's contract on tool names. A skill names the need and the acceptance criteria. A platform tool is named only where calling it is the goal of the step — set the branding, register a webhook, flip a workspace flag — and never where it's merely one way to fetch data. So "the questions readers asked that got no answer" stays a need: an agent with nothing connected greps the logs, and one with the MCP connected gets the same answer as a typed list. That's what makes a skill run on a bare agent and get sharper — not narrower — when a platform is attached.

Modes — what a skill is allowed to touch

Every skill declares a mode, so you know before you hand it your docs whether it edits files:

| Mode | Reads your docs | Writes files | What it's for | |---|---|---|---| | audit | yes | never | Reports findings you decide what to do with | | refactor | yes | edits pages that already exist | Fixes; never invents a new page | | authoring | — | writes new pages | Rules loaded before you draft, so the page ships correct | | platform | — | workspace settings, not content | Branding, languages, webhooks, domains |

CI enforces it: a skill without a valid mode does not ship.

| Concept | Analogy | Lives in | |---|---|---| | Skill | QA checklist — Workflow, Guardrails, Acceptance Criteria | this repo | | Subagent | Jira ticket — specific model, tools, event trigger | docs-subagents |

Optional acceleration

| Path | What you get | Cost | |---|---|---| | Nothing (default) | grep/find over the docs folder | free, works everywhere | | markdown-lsp (self-hosted) | semantic + graph search, runs locally — repo | free, you host it | | Docsbook MCP (cloud) | the same search in the cloud, plus workspace settings | optional account |

# Optional cloud transport
mcp add --transport http https://docsbook.io/api/mcp/server

The cloud transport runs markdown-lsp for you — self-hosted and cloud are the same capability, your choice of where it runs. The bare grep/find fallback always works.

CLI reference

Powered by the skills CLI:

npx skills add Docsbook-io/docs-skills --skill '*'        # install the whole catalog
npx skills add Docsbook-io/docs-skills --skill docs-seo   # install one skill
npx skills add Docsbook-io/docs-skills -a claude-code -a cursor --skill '*'   # target specific agents
npx skills list                                           # list installed skills
npx skills find <keyword>                                 # search for a skill
npx skills update                                         # update installed skills

Contributing

Skills are plain Markdown files under skills/ — either skills/<name>/SKILL.md or skills/<category>/<name>/SKILL.md; the category comes from frontmatter, not the path. To add one:

1. Create skills/your-skill/SKILL.md following the schema in schema/
2. Run pnpm build-index to regenerate index.json
3. Open a PR

The catalog is intentionally minimal — one file per skill, no runtime dependencies.

License

MIT © 2024 Dan Bondarev / docsbook.io