Use typed-openapi with an AI agent

The repository includes a skill file for coding agents. It tells an agent where generated code lives, how to choose generation options, and what to verify after a change.

Copy the skill into the agent’s project-specific skills location, or attach it directly when your agent supports file context. Keep the OpenAPI document, typed-openapi.config.ts, and generated output in the agent’s scope.

Paste this prompt

Use the typed-openapi skill in this repository.

I need to integrate the OpenAPI document at <path-to-openapi.yaml> into <app/package/path>.
First inspect the existing typed-openapi config and generated files. Do not hand-write endpoint types or edit generated files directly.

Choose the smallest setup that fits this app:
- keep types-only output unless runtime validation is already needed;
- use the generated default fetcher only after checking each operation's OpenAPI security requirements and showing its base-URL design;
- reuse the app's existing HTTP layer when it already owns retry, auth, tracing, or environment access;
- generate TanStack Query or MSW output only when this app uses those libraries.

For browser apps, use the generated `createApi(baseUrl)` factory instead of editing generated code. For query API keys, verify each operation's security requirements and add a focused test that confirms the key and ordinary query parameters both reach the server.

Show the exact config and command before changing files. Then regenerate, run the relevant typecheck/tests, and summarize the generated files and any OpenAPI gaps you found.

The prompt is intentionally specific about generated files. It stops an agent from quietly replacing a transport policy or creating a second hand-written client.