posthog/ai-plugin

building-html-canvases

Author a PostHog canvas with semantic HTML, CSS, and direct browser APIs — documents, articles, generative graphics, 2D canvas and WebGL experiences, and focused experiments where React components add no useful structure.

ソースを見る

使い方

  1. インストールコマンドをコピーし、自分の環境へ導入する前にソースを確認します。
  2. 必要なツール、認証情報、対応環境を原文で確認します。
  3. 入力と成果物を指定し、出力を資料と照合します。
npx skills add https://github.com/posthog/ai-plugin --skill building-html-canvases

開始前の確認

  • 導入にはモデル契約、API枠、ファイルへのアクセス許可は含まれません。
  • スター数や導入数は活動の指標であり、独立した品質・安全性試験ではありません。
リポジトリの原文

見出し、例、コード、表、リンク、参照画像を含む原文を表示しています。

Building HTML canvases

Some canvases are documents or graphics programs, not applications: a written report, a diagram, a generative-art piece, a WebGL scene. For these, semantic HTML, CSS, and direct browser APIs are the right tools — don't force Quill components or React state onto a static page.

The wrapper the current runtime requires

Every canvas keeps src/canvas.tsx as its mounted React entry component (default export, no props). Keep the React layer as a thin shell and write the experience in HTML/CSS/browser APIs inside it:

  • A document is JSX that is effectively semantic HTML — <article>, headings, lists, tables,

figures — with a <style> block for typography and layout. Write real, specific copy.

  • A drawing/WebGL program renders a <canvas> element and drives it imperatively from a

useEffect via a ref: get the 2D/WebGL context, run the setup and render loop there.

  • Clean up in the effect's return: cancel requestAnimationFrame loops, remove listeners, and

release contexts, so theme switches and remounts don't leak or double-run.

  • Mixing tiers is fine: a mostly static page can mount one interactive island, and a data board

can hand a chart's <canvas> to imperative code while React owns the chrome.

The import allowlist still applies (react, react-dom, @posthog/quill, recharts, lucide-react, dayjs) — browser globals (document, CanvasRenderingContext2D, WebGLRenderingContext, requestAnimationFrame, IntersectionObserver, Web Audio, etc.) need no import. Three.js and other npm graphics libraries are not yet loadable; write against raw WebGL or 2D canvas until the build pipeline's dependency admission ships.

Styling and theme without Quill

  • Size the outermost JSX/HTML element to the iframe viewport with h-screen or height: 100vh.

Do not use h-full or height: 100% on that root: a published canvas's artifact shell gives its html, body, and #root elements no explicit height, so percentage height collapses to the content height. Descendants may use percentage height after the outermost element establishes the viewport height.

  • Use Tailwind utilities and/or a <style> block (keyframes and complex selectors are fine).
  • The host toggles a .dark class on the document root when the user's PostHog theme changes.

Define your colors as CSS variables under :root { … } with overrides under html.dark { … }, or use theme token utilities (bg-background, text-foreground, border-border) — never a light-only hardcoded color.

  • For canvas/WebGL drawing colors, read the resolved token at runtime

(getComputedStyle(document.documentElement).getPropertyValue("--primary")) or your own CSS variables, and re-read on theme change if the scene is long-lived.

Rules that still apply

  • PostHog data comes only through the ph bridge (see querying-canvas-data), including

ph.capture for interaction analytics. Other requests and external styles, images, fonts, media, or frames require their exact public HTTPS origins in capabilities.network.origins and work only after publishing. Remote scripts and dynamic imports remain blocked.

  • A document that states PostHog numbers must make each one verifiable: an insight-backed number

links its saved insight through ph.openExternal (URL from the generate-app-url MCP tool, from a click); an ad-hoc ph.query number discloses the exact query that ran in a <details> element beside the claim — see "Verifiability" in querying-canvas-data.

  • External links go through ph.openExternal(url) (posthog.com origins only), from a user

interaction.

  • Validate and publish through the canvas tools as described in validating-and-publishing-canvases.

タスクに関連する項目

同じリポジトリから

関連する Skills

すべての Skills
posthog
公式

composing-grid-canvases

Compose PostHog grid canvases — widget grids (including the user's home canvas) built from reusable component canvases. Use when a task asks to add, fill, move, resize, or remove a widget on a grid or home canvas, to compose a whole canvas of widgets from one ask, to build a reusable widget/component, or when a placement id or grid canvas id is the target. Covers the component store search → configure → fork → build ladder, the component placement contract (size, configSchema), the placement lif

1時間の導入数
1
GitHub Stars
96
更新日
10月31日
posthog
公式

consuming-endpoints-from-client-code

Wire a PostHog endpoint into a client app or SDK. Covers fetching the OpenAPI spec, generating a typed client with openapi-generator or @hey-api/openapi-ts, sending the right auth header, shaping the variables payload (HogQL codename vs insight breakdown property), handling rate-limit and materialised-endpoint error responses. Use when the user says "how do I call my endpoint", "generate a client for this", or "what auth header do I use".

1時間の導入数
1
GitHub Stars
96
更新日
10月31日
posthog
公式

designing-email-templates

Author, save, and edit email templates in the PostHog workflows library — compose email design JSON with Liquid personalization and create and round-trip-edit templates over MCP. Use when asked to design, build, update, or fix an email template for workflows, broadcasts, or campaigns.

1時間の導入数
1
GitHub Stars
96
更新日
10月31日
posthog
公式

exploring-llm-evaluations

Investigate AI observability evaluations — hog (deterministic code-based), llmjudge (LLM-prompt-based), and sentiment (user-message sentiment). Find existing evaluations, inspect their configuration, run them against specific generations, query individual results, and set up scheduled reports on an evaluation. Use when the user asks to debug why an evaluation is failing, surface common failure modes, compare results across filters, dry-run a Hog evaluator, prototype a new LLM-judge prompt, inspe

1時間の導入数
1
GitHub Stars
96
更新日
10月31日