Cloudflare Doctor

Use this skill to audit a user's Cloudflare project like a doctor: diagnose from evidence, name the risk precisely, explain the Cloudflare primitive/product mismatch, and prescribe the smallest safe fix.

First principles

• Evidence first: every finding needs a file/config/account source, line/reference where possible, and a concrete Cloudflare behavior or billing mechanism.
• Current docs over memory: before asserting a Cloudflare best practice, product behavior, limit, pricing detail, or configuration recommendation, fetch the current official Cloudflare docs. Never rely on the agent's training data or memory when live docs can be fetched. If docs cannot be fetched, say the claim was not current-doc verified.
• Do not infer dashboard/account settings from repo files. If DNS, SSL/TLS, WAF, cache rules, Access, billing, or Logpush evidence is absent, ask for Terraform/export/screenshots/API output or mark it not inspected.
• Separate confirmed findings from suspicions to verify. Cloudflare products and limits change; cite current docs/pricing when making date-sensitive claims.
• Prefer primitive-fit fixes over local patches: KV vs D1 vs Durable Objects vs R2 vs Queues vs Workflows vs Cache is often the root problem.
• Cost issues are findings even when the app works. Estimate the cost mechanism, not exact dollars unless the user supplied volumes and plan details.

Standard workflow