Cloudflare Deploy: Comprehensive Edge Platform Deployment
The Cloudflare deploy skill from OpenAI provides consolidated deployment capabilities across Cloudflare's entire platform—Workers, Pages, Durable Objects, D1, R2, and dozens of other services. Rather than requiring knowledge of each product's deployment nuances, this skill uses decision trees and product references to guide appropriate service selection and deployment configuration.
What This Skill Does
This skill enables AI assistants to deploy applications and infrastructure to Cloudflare's edge platform through a unified interface. It handles authentication verification, service selection through decision trees, deployment execution with appropriate permissions, and troubleshooting common issues like network access requirements.
The skill organizes Cloudflare's extensive product catalog into logical categories: compute and runtime (Workers, Pages, Durable Objects), storage and data (KV, D1, R2, Queues), AI and machine learning (Workers AI, Vectorize, Agents SDK), networking and connectivity (Tunnel, Spectrum), security (WAF, DDoS, Turnstile), media and content (Images, Stream), and infrastructure-as-code (Pulumi, Terraform).
Decision trees help navigate product selection. Need to run code? The tree branches to Workers for serverless functions, Pages for full-stack apps, Durable Objects for stateful coordination, or Workflows for long-running jobs. Need storage? It directs to KV for key-value, D1 for SQL, R2 for objects, Queues for messages, or Vectorize for vector embeddings. This guided approach prevents choice paralysis when facing 40+ products.
Getting Started
Before deploying anything, verify authentication using npx wrangler whoami. This shows account information if authenticated. If not authenticated, the skill loads appropriate authentication references—interactive environments use wrangler login for one-time OAuth, while CI/CD environments set CLOUDFLARE_API_TOKEN environment variables.
When sandbox networking blocks deployment calls, the skill reruns with escalated permissions using sandbox_permissions=require_escalated. This is common because Cloudflare deployments require outbound network access that may be restricted in sandboxed environments.
Deployments can take several minutes. The skill uses appropriate timeout values to prevent premature termination, especially for complex applications or large asset uploads during Pages deployments.
Key Features
Unified Decision Trees: Instead of requiring users to know which Cloudflare product they need, decision trees ask capability-oriented questions. "Need to run code?" "Need storage?" "Need AI?" Based on answers, the skill identifies appropriate products and loads detailed references.
Comprehensive Product Coverage: The skill supports deployment to Workers (serverless functions), Pages (static sites with functions), Durable Objects (stateful edge objects), D1 (SQLite at edge), R2 (S3-compatible storage), Queues (message queues), Workers AI (inference), Vectorize (vector database), and dozens more services.
Authentication Handling: Before deployment, the skill verifies authentication status and loads appropriate setup references when authentication is missing. This prevents failed deployments due to credential issues.
Network Permission Management: When sandbox networking blocks deployments, the skill automatically requests escalated permissions and reruns commands. This handles a common deployment friction point transparently.
Product-Specific References: Each Cloudflare product has dedicated reference documentation loaded on demand. Once decision trees identify the right product, detailed guides ensure correct deployment configuration.
Usage Examples
When a user wants to deploy a serverless API, the skill asks about requirements—serverless functions at the edge points to Workers. It verifies authentication, loads Workers references, configures wrangler.toml, and executes wrangler deploy with appropriate timeouts and network permissions.
For deploying a full-stack web application connected to a Git repository, decision trees identify Pages as appropriate. The skill sets up Pages deployment with Git integration, configures build settings, handles environment variables, and deploys with wrangler pages deploy.
When building AI-powered applications requiring vector search, the skill recommends Workers AI for inference and Vectorize for vector database capabilities. It configures bindings between services, sets up storage, and ensures deployed Workers can access AI models and vector indexes.
For projects needing object storage, the skill identifies R2 as S3-compatible storage solution. It configures buckets, sets up bindings to Workers or Pages Functions, handles CORS if needed, and ensures applications can read/write objects at edge locations.
Best Practices
Always verify authentication before deployment attempts. Running npx wrangler whoami takes seconds and prevents failed deployments due to missing credentials. Make authentication verification the first step of any deployment workflow.
Use decision trees rather than jumping directly to specific products. Even when you think you know the right service, walking through decision trees might reveal better-suited alternatives. Cloudflare's product catalog is extensive—guided selection prevents suboptimal choices.
Set appropriate timeout values for deployments. Complex applications, large asset uploads, or first-time deployments can take minutes. Default timeouts may be too short. The skill uses generous timeouts to prevent premature termination.
Request escalated permissions proactively when deploying from sandboxed environments. Rather than waiting for network errors, request escalated permissions at deployment initiation. This prevents failed attempts and retries.
Combine products appropriately for complete solutions. Most real applications need multiple Cloudflare services—Workers plus D1 for database, Pages plus R2 for storage, Workers AI plus Vectorize for AI features. The skill helps architect these combinations through its decision trees.
When to Use This Skill
Use this skill when deploying applications or infrastructure to Cloudflare's edge platform. Whether building serverless APIs, full-stack web apps, AI-powered services, or media processing pipelines, this skill handles the deployment complexity.
The skill is particularly valuable when uncertain which Cloudflare products fit your use case. Decision trees guide selection based on capabilities rather than requiring product knowledge upfront. This makes Cloudflare accessible even when unfamiliar with its extensive catalog.
It's ideal for automated deployment workflows. CI/CD pipelines benefit from the skill's authentication handling, network permission management, and comprehensive product support. Deployments become reproducible and reliable.
When NOT to Use This Skill
Don't use this skill for deploying to other cloud platforms. It's specific to Cloudflare's edge platform. For AWS, Azure, or Google Cloud deployments, use appropriate platform-specific skills.
Avoid using it when Cloudflare's edge architecture doesn't fit requirements. Some applications need long-running servers, persistent connections, or capabilities better suited to traditional cloud infrastructure. Edge platforms have constraints—request/response time limits, cold start considerations, regional data requirements.
It's not appropriate for managing existing Cloudflare resources beyond deployment. While the skill handles deployment, comprehensive resource management—monitoring, analytics, configuration updates, access control—requires broader Cloudflare administration capabilities.
Don't expect the skill to make architectural decisions about which products to combine. While decision trees guide individual product selection, architecting complete solutions requires understanding application requirements and making informed tradeoffs.
Related Skills
This skill complements vercel-deploy for Vercel deployments, netlify-deploy for Netlify deployments, and render-deploy for Render deployments across different edge and cloud platforms.
Source
This skill is maintained by OpenAI. View on GitHub