Orange Book
Guides

Launch checklist for AI-generated code

Move an AI-generated Cloudflare project from “looks runnable” to verified APIs, permissions, tests, cost, and rollback.

Edited and verified by Orange Book Editorial Team ·

AI BUILDERS · RELEASE CHECK25 minutesVerified 2026-08-26

An AI completion claim is not acceptance evidence

“Fixed,” “deployed,” and “Workers-compatible” require a command, response, or current primary-source API page. An assistant cannot know your final Dashboard state without evidence you provide.

Six gates

GateEvidence to retain
APIsA current official source for every Wrangler field and Runtime API
AuthorizationMinimum token, resource scope, and clean secret scan
CorrectnessSuccess, invalid input, 404, and downstream-failure tests
RuntimeA real workerd or preview result
CostGzip bundle size, CPU, request, and downstream-operation model
RollbackKnown-good version, data compatibility, and trigger conditions

Execution order

Read configuration instead of guessing

Check each field against the current Wrangler reference. Verify compatibility date, bindings, and environments. Remove real IDs, tokens, and broad permissions generated by AI.

Search for server assumptions

Inspect filesystem, resident memory, listening sockets, native modules, and Node stubs. Use Runtime versus Node.js.

Run failure paths

Cover empty and malicious input, missing resources, absent bindings, and downstream failure. Return stable client errors without stack traces.

Build and exercise workerd

Run lint, types, and tests, then a production build and Wrangler dry run. Finally observe status, headers, and real data boundaries at a preview URL.

Review change and rollback

Read the diff manually and list architecture changes and irreversible actions. D1 migrations, DNS, WAF, and published routes always require separate confirmation.

Primary sources

Did this page help you complete your goal?

Beta feedback is generated in this browser and is never uploaded automatically.

On this page