Cloudflare Workers Platform Skill

Consolidated skill for building on the Cloudflare platform. Biases toward retrieval from live Cloudflare docs over pre-trained knowledge - API signatures, limits, and pricing change frequently.

Step 1: Choose the Right Product

"I need to run code"

"I need to store data"

"I need AI/ML"

"I need networking/security"

Step 2: Workers Fundamentals

Minimal Worker (TypeScript)

export default {
  async fetch(request: Request, env: Env, ctx: ExecutionContext): Promise<Response> {
    const url = new URL(request.url)

    if (url.pathname === '/health') {
      return Response.json({ status: 'ok' })
    }

    return new Response('Hello, World!', { status: 200 })
  },
} satisfies ExportedHandler<Env>

interface Env {
  MY_KV: KVNamespace
  MY_DB: D1Database
  MY_BUCKET: R2Bucket
  SECRET_KEY: string
}

wrangler.toml Configuration

name = "my-worker"
main = "src/index.ts"
compatibility_date = "2025-04-10"
compatibility_flags = ["nodejs_compat"]

[[kv_namespaces]]
binding = "MY_KV"
id = "xxxxxxxxxxxxxxxx"

[[d1_databases]]
binding = "MY_DB"
database_name = "my-database"
database_id = "xxxxxxxxxxxxxxxx"

[[r2_buckets]]
binding = "MY_BUCKET"
bucket_name = "my-bucket"

[ai]
binding = "AI"

Essential Wrangler Commands

npm create cloudflare@latest   # scaffold a new project
wrangler dev                   # local dev server (http://localhost:8787)
wrangler deploy                # deploy to production
wrangler tail                  # stream real-time logs
wrangler secret put SECRET_KEY # add encrypted secret
wrangler kv key put --binding MY_KV "key" "value"

Step 3: KV Storage

// Write
await env.MY_KV.put('user:123', JSON.stringify({ name: 'Alice' }), {
  expirationTtl: 3600  // seconds
})

// Read
const raw = await env.MY_KV.get('user:123')
const user = raw ? JSON.parse(raw) : null

// Delete
await env.MY_KV.delete('user:123')

// List keys
const { keys } = await env.MY_KV.list({ prefix: 'user:' })

KV Characteristics:

• Eventually consistent (changes propagate in ~60s globally)
• Read-optimized (millions of reads/s, ~1 write/s per key)
• Not suitable for high-frequency writes - use Durable Objects for that

Step 4: D1 Database (SQL)

// Query
const { results } = await env.MY_DB.prepare(
  'SELECT * FROM users WHERE email = ?'
).bind('[email protected]').all()

// Insert
await env.MY_DB.prepare(
  'INSERT INTO users (name, email) VALUES (?, ?)'
).bind('Alice', '[email protected]').run()

// Batch operations
await env.MY_DB.batch([
  env.MY_DB.prepare('UPDATE users SET active = 1 WHERE id = ?').bind(1),
  env.MY_DB.prepare('INSERT INTO logs (action) VALUES (?)').bind('activated'),
])

Schema migrations - use wrangler d1 migrations:

wrangler d1 migrations create my-database add-users-table
wrangler d1 migrations apply my-database --local   # local
wrangler d1 migrations apply my-database           # production

Step 5: R2 Object Storage

// Upload
await env.MY_BUCKET.put('files/photo.jpg', request.body, {
  httpMetadata: { contentType: 'image/jpeg' },
})

// Download
const object = await env.MY_BUCKET.get('files/photo.jpg')
if (!object) return new Response('Not Found', { status: 404 })
return new Response(object.body, {
  headers: { 'Content-Type': object.httpMetadata?.contentType ?? 'application/octet-stream' },
})

// Delete
await env.MY_BUCKET.delete('files/photo.jpg')

// List objects
const listed = await env.MY_BUCKET.list({ prefix: 'files/', limit: 100 })

Step 6: Workers AI

// Text generation
const response = await env.AI.run('@cf/meta/llama-3.1-8b-instruct', {
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'Explain DNS in one paragraph.' },
  ],
})
return Response.json({ text: response.response })

// Embeddings
const embeds = await env.AI.run('@cf/baai/bge-small-en-v1.5', {
  text: ['Hello world', 'Goodbye world'],
})
// embeds.data is an array of float arrays

// Image classification
const result = await env.AI.run('@cf/microsoft/resnet-50', {
  image: [...new Uint8Array(await request.arrayBuffer())],
})

Available models: check https://developers.cloudflare.com/workers-ai/models/

Step 7: Durable Objects

// src/counter.ts - the Durable Object class
export class Counter implements DurableObject {
  state: DurableObjectState
  value: number = 0

  constructor(state: DurableObjectState, env: Env) {
    this.state = state
    this.state.blockConcurrencyWhile(async () => {
      this.value = (await this.state.storage.get<number>('value')) ?? 0
    })
  }

  async fetch(request: Request): Promise<Response> {
    const url = new URL(request.url)
    if (url.pathname === '/increment') {
      this.value++
      await this.state.storage.put('value', this.value)
    }
    return Response.json({ value: this.value })
  }
}

// src/index.ts - Worker that uses it
export default {
  async fetch(request: Request, env: Env) {
    const id = env.COUNTER.idFromName('global')
    const stub = env.COUNTER.get(id)
    return stub.fetch(request)
  },
}

interface Env {
  COUNTER: DurableObjectNamespace
}

wrangler.toml:

[[durable_objects.bindings]]
name = "COUNTER"
class_name = "Counter"

[[migrations]]
tag = "v1"
new_classes = ["Counter"]

Step 8: Queues (Background Processing)

// Producer - enqueue from a Worker
await env.MY_QUEUE.send({ userId: 123, action: 'send-welcome-email' })

// Consumer - process messages
export default {
  async queue(batch: MessageBatch<{ userId: number; action: string }>, env: Env) {
    for (const msg of batch.messages) {
      await processMessage(msg.body)
      msg.ack()  // acknowledge on success
    }
  },
} satisfies ExportedHandler<Env>

Step 9: Cloudflare Pages

Deploy full-stack apps with Git integration:

npm create cloudflare@latest my-app -- --framework=next
cd my-app
wrangler pages deploy .next   # or use the dashboard for Git integration

Pages Functions (serverless backend):

// functions/api/user.ts
export async function onRequest(ctx: EventContext<Env, '/api/user', {}>) {
  return Response.json({ user: 'Alice' })
}

Common Mistakes

• Accessing live Cloudflare docs for exact limits - free/paid tier limits change; check https://developers.cloudflare.com/
• Using KV for high-write workloads - KV is eventually consistent and write-limited; use Durable Objects instead
• Missing `compatibility_date` - always set to a recent date to get latest APIs
• Blocking the event loop - Workers are single-threaded; avoid CPU-heavy synchronous work
• Storing secrets in `wrangler.toml` - use wrangler secret put for sensitive values
• Using `nodejs` APIs without `nodejs_compat` flag - add compatibility_flags = ["nodejs_compat"] in wrangler.toml