Get cloudflare skills from middlecache

.gitignore

@@ -5,6 +5,9 @@ distllms/
 # generated types
 .astro/
 
+# skills/ is fetched from middlecache via bin/fetch-skills.ts
+skills/
+
 # dependencies
 node_modules/
 

AGENTS.md

@@ -35,7 +35,11 @@
 ├── public/                 # Static files served as-is (images, redirects, robots.txt)
 ├── worker/                 # Cloudflare Worker for serving the site
 ├── bin/                    # Build scripts and CI helpers
-├── skills/                 # Interactive exercises (astro-skills)
+│   └── fetch-skills.ts     # Downloads skills.tar.gz from middlecache, extracts to skills/
+├── skills/                 # Agent Skills served at /.well-known/skills/ — GENERATED, do not edit
+│                           # Fetched from https://middlecache.ced.cloudflare.com/v1/cloudflare-skills/skills.tar.gz
+│                           # by bin/fetch-skills.ts, which runs automatically via prebuild/predev hooks.
+│                           # skills/ is in .gitignore and is NOT committed to the repository.
 ├── astro.config.ts         # Astro + Starlight configuration
 ├── ec.config.mjs           # Expressive Code (syntax highlighting) configuration
 ├── package.json
@@ -63,19 +67,19 @@
 
 ```yaml
 ---
-title: Page Title                    # Required
-description: SEO meta description    # Recommended
-pcx_content_type: how-to             # Page type (see below)
+title: Page Title # Required
+description: SEO meta description # Recommended
+pcx_content_type: how-to # Page type (see below)
 sidebar:
-  order: 1                           # Sort order in sidebar
-  label: Custom Label                # Override sidebar text
-tags:                                # Optional, validated against allowlist
+  order: 1 # Sort order in sidebar
+  label: Custom Label # Override sidebar text
+tags: # Optional, validated against allowlist
   - JavaScript
   - Workers
-products:                            # References to src/content/products/ entries
+products: # References to src/content/products/ entries
   - workers
-difficulty: Beginner                 # For tutorials: Beginner | Intermediate | Advanced
-reviewed: 2025-01-15                 # YYYY-MM-DD of last content review
+difficulty: Beginner # For tutorials: Beginner | Intermediate | Advanced
+reviewed: 2025-01-15 # YYYY-MM-DD of last content review
 ---
 ```
 
@@ -87,10 +91,10 @@
 
 MDX is parsed as JSX, not plain Markdown. These characters have special meaning and **will break the build** if used unescaped in prose:
 
-| Character | Problem | Fix |
-|-----------|---------|-----|
-| `{` `}` | Interpreted as JS expressions | Wrap in backticks or use `\{` `\}` |
-| `<` `>` | Interpreted as JSX elements | Use `&lt;` `&gt;` or wrap in backticks |
+| Character | Problem                       | Fix                                    |
+| --------- | ----------------------------- | -------------------------------------- |
+| `{` `}`   | Interpreted as JS expressions | Wrap in backticks or use `\{` `\}`     |
+| `<` `>`   | Interpreted as JSX elements   | Use `&lt;` `&gt;` or wrap in backticks |
 
 This is the single most common build failure. Always check prose, tables, and headings for these characters.
 
@@ -130,7 +134,14 @@
 Components are imported from `~/components` in MDX files:
 
 ```mdx
-import { Render, TypeScriptExample, WranglerConfig, Details } from "~/components";
+import {
+	Render,
+	TypeScriptExample,
+	WranglerConfig,
+	Details,
+} from "~/components";
+
+;
 ```
 
 Components **must** be imported after the frontmatter block. Forgetting the import is a common mistake.
@@ -143,6 +154,7 @@
 <Render file="partial-name" product="workers" />
 
 <!-- With parameters: -->
+
 <Render file="partial-name" product="workers" params={{ key: "value" }} />
 ```
 
@@ -164,14 +176,15 @@
 
 Shows Wrangler configuration in both TOML and JSON formats with synced tabs. Auto-converts between formats.
 
-```mdx
+````mdx
 <WranglerConfig>
 
 ```toml
 name = "my-worker"
 main = "src/index.ts"
 compatibility_date = "$today"
 ```
+````
 
 </WranglerConfig>
 ```
@@ -185,8 +198,8 @@
 import { Tabs, TabItem } from "~/components";
 
 <Tabs>
-  <TabItem label="npm">npm install package</TabItem>
-  <TabItem label="yarn">yarn add package</TabItem>
+	<TabItem label="npm">npm install package</TabItem>
+	<TabItem label="yarn">yarn add package</TabItem>
 </Tabs>
 ```
 
@@ -212,19 +225,19 @@
 
 ### Other frequently used components
 
-| Component | Purpose |
-|-----------|---------|
-| `Plan` | Display plan availability (e.g., `<Plan type="enterprise" />`) |
-| `GlossaryTooltip` | Inline hover tooltip with glossary definition |
-| `InlineBadge` | Status badges: `<InlineBadge preset="beta" />` |
-| `LinkTitleCard` | Navigation card with icon, title, and description |
-| `DirectoryListing` | Auto-generated listing of child pages |
-| `YouTube` | Embed YouTube video by ID |
-| `Stream` | Embed Cloudflare Stream video |
-| `APIRequest` | Generate curl commands from the Cloudflare OpenAPI schema |
-| `DashButton` | "Go to Dashboard" button with validated deeplink |
-| `ListTutorials` | Auto-generated tutorial listing table |
-| `GitHubCode` | Fetch and display code from a GitHub repository |
+| Component          | Purpose                                                        |
+| ------------------ | -------------------------------------------------------------- |
+| `Plan`             | Display plan availability (e.g., `<Plan type="enterprise" />`) |
+| `GlossaryTooltip`  | Inline hover tooltip with glossary definition                  |
+| `InlineBadge`      | Status badges: `<InlineBadge preset="beta" />`                 |
+| `LinkTitleCard`    | Navigation card with icon, title, and description              |
+| `DirectoryListing` | Auto-generated listing of child pages                          |
+| `YouTube`          | Embed YouTube video by ID                                      |
+| `Stream`           | Embed Cloudflare Stream video                                  |
+| `APIRequest`       | Generate curl commands from the Cloudflare OpenAPI schema      |
+| `DashButton`       | "Go to Dashboard" button with validated deeplink               |
+| `ListTutorials`    | Auto-generated tutorial listing table                          |
+| `GitHubCode`       | Fetch and display code from a GitHub repository                |
 
 For the full component list and their props, see `src/components/index.ts` (barrel export) and the individual `.astro` / `.tsx` files.
 
@@ -294,7 +307,7 @@
 6. Redirect validation (`bin/validate-redirects.ts`)
 7. `npm run test` (all Vitest suites)
 
 A separate Semgrep workflow checks style guide compliance (dates, "coming soon" phrases) and produces warnings.
 
 ## Common mistakes to avoid
 
@@ -315,27 +328,27 @@
 
 The site defines 20 content collections in `src/content.config.ts` with schemas in `src/schemas/`. The major ones:
 
-| Collection | Location | Description |
-|-----------|----------|-------------|
-| `docs` | `src/content/docs/` | Main documentation pages (MDX) |
-| `partials` | `src/content/partials/` | Reusable content snippets (MDX) |
-| `changelog` | `src/content/changelog/` | Product changelogs (MDX) |
-| `glossary` | `src/content/glossary/` | Glossary terms (YAML) |
-| `products` | `src/content/products/` | Product metadata (YAML) |
-| `plans` | `src/content/plans/` | Plan/pricing data (YAML) |
-| `workers-ai-models` | `src/content/workers-ai-models/` | AI model definitions (JSON) |
-| `fields` | `src/content/fields/` | Ruleset engine field definitions (YAML) |
-| `learning-paths` | `src/content/learning-paths/` | Learning path definitions (JSON) |
+| Collection          | Location                         | Description                             |
+| ------------------- | -------------------------------- | --------------------------------------- |
+| `docs`              | `src/content/docs/`              | Main documentation pages (MDX)          |
+| `partials`          | `src/content/partials/`          | Reusable content snippets (MDX)         |
+| `changelog`         | `src/content/changelog/`         | Product changelogs (MDX)                |
+| `glossary`          | `src/content/glossary/`          | Glossary terms (YAML)                   |
+| `products`          | `src/content/products/`          | Product metadata (YAML)                 |
+| `plans`             | `src/content/plans/`             | Plan/pricing data (YAML)                |
+| `workers-ai-models` | `src/content/workers-ai-models/` | AI model definitions (JSON)             |
+| `fields`            | `src/content/fields/`            | Ruleset engine field definitions (YAML) |
+| `learning-paths`    | `src/content/learning-paths/`    | Learning path definitions (JSON)        |
 
 ## Testing
 
 Tests use Vitest with three workspace projects (`vitest.workspace.ts`):
 
-| Suite | File pattern | Runtime |
-|-------|-------------|---------|
+| Suite   | File pattern       | Runtime                           |
+| ------- | ------------------ | --------------------------------- |
 | Workers | `*.worker.test.ts` | `@cloudflare/vitest-pool-workers` |
-| Node | `*.node.test.ts` | Node.js |
-| Astro | `*.astro.test.ts` | Astro Vite config |
+| Node    | `*.node.test.ts`   | Node.js                           |
+| Astro   | `*.astro.test.ts`  | Astro Vite config                 |
 
 Run all tests: `npm run test`
 

bin/fetch-skills.ts

@@ -0,0 +1,89 @@
+#!/usr/bin/env tsx
+
+import { spawn } from "child_process";
+import fs from "fs";
+import { join } from "path";
+
+import { downloadToDotTempIfNotPresent } from "../src/util/custom-loaders";
+
+const MIDDLECACHE_BASE_URL = "https://middlecache.ced.cloudflare.com/";
+const SKILLS_MIDDLECACHE_PATH = "v1/cloudflare-skills/skills.tar.gz";
+const SKILLS_DOT_TMP_PATH = `middlecache/${SKILLS_MIDDLECACHE_PATH}`;
+const MANIFEST_MIDDLECACHE_PATH = "v1/cloudflare-skills/skills-manifest.json";
+const MANIFEST_DOT_TMP_PATH = `middlecache/${MANIFEST_MIDDLECACHE_PATH}`;
+const SKILLS_DIR = "./skills";
+
+// --soft: warn and continue on failure instead of exiting non-zero.
+//         Used by the predev hook so a network failure doesn't block local development.
+// --force: re-fetch even if skills/ already exists.
+const soft = process.argv.includes("--soft");
+const force = process.argv.includes("--force");
+
+const fail = (message: string): never => {
+	if (soft) {
+		const hasExisting = fs.existsSync(SKILLS_DIR);
+		console.warn(
+			hasExisting
+				? `Warning: ${message} — continuing with existing Cloudflare Skills`
+				: `Warning: ${message} — skills/ does not exist, /.well-known/skills/ will not work`,
+		);
+		process.exit(0);
+	}
+	console.error(`Error: ${message}`);
+	process.exit(1);
+};
+
+if (fs.existsSync(SKILLS_DIR) && !force) {
+	console.log(
+		"/skills directory already exists, skipping fetch. (run `npx tsx bin/fetch-skills.ts --force` to re-fetch)",
+	);
+	process.exit(0);
+}
+
+console.log("Fetching Cloudflare Skills from middlecache");
+
+try {
+	await Promise.all([
+		downloadToDotTempIfNotPresent(
+			`${MIDDLECACHE_BASE_URL}${SKILLS_MIDDLECACHE_PATH}`,
+			SKILLS_DOT_TMP_PATH,
+		),
+		downloadToDotTempIfNotPresent(
+			`${MIDDLECACHE_BASE_URL}${MANIFEST_MIDDLECACHE_PATH}`,
+			MANIFEST_DOT_TMP_PATH,
+		),
+	]);
+} catch (err) {
+	fail(`fetch failed: ${err}`);
+}
+
+const tarballPath = join(".tmp", ...SKILLS_DOT_TMP_PATH.split("/"));
+
+// Remove existing skills/ directory so stale Cloudflare Skills don't accumulate
+fs.rmSync(SKILLS_DIR, { recursive: true, force: true });
+fs.mkdirSync(SKILLS_DIR, { recursive: true });
+
+// Extract the tarball from .tmp/ into ./skills/.
+// The archive contains skills/<skill-name>/... so we strip the leading "skills/"
+// component and extract into SKILLS_DIR.
+const tar = spawn(
+	"tar",
+	["--strip-components=1", "-xz", "-C", SKILLS_DIR, "-f", tarballPath],
+	{ stdio: "inherit" },
+);
+
+const exitCode = await new Promise<number | null>((resolve) =>
+	tar.on("close", resolve),
+);
+
+if (exitCode !== 0) {
+	fail(`tar exited with code ${exitCode}`);
+}
+
+const cloudflareSkills = fs
+	.readdirSync(SKILLS_DIR)
+	.filter((entry) => fs.statSync(`${SKILLS_DIR}/${entry}`).isDirectory());
+
+console.log(
+	`Fetched ${cloudflareSkills.length} Cloudflare Skills: ${cloudflareSkills.join(", ")}`,
+);

package.json

@@ -4,11 +4,13 @@
 	"type": "module",
 	"scripts": {
 		"astro": "npx astro",
+		"prebuild": "npx tsx bin/fetch-skills.ts",
 		"build": "export NODE_OPTIONS='--max-old-space-size=6192' || set NODE_OPTIONS=\"--max-old-space-size=6192\" && npx astro build",
 		"typegen:worker": "npx wrangler types ./worker/worker-configuration.d.ts",
 		"check": "npm run check:astro && npm run check:worker",
 		"check:astro": "npx astro check",
 		"check:worker": "npx tsc --noEmit -p ./worker/tsconfig.json",
+		"predev": "npx tsx bin/fetch-skills.ts --soft",
 		"dev": "npx astro dev",
 		"format": "npm run format:core:fix && npm run format:data",
 		"format:core": "npx prettier \"**/*.{js,jsx,ts,tsx,mjs,css}\"",