mcpbeat

Integration Tests

openai/integration-tests

Run the integration-tests pipeline that depends on a local npm registry (Verdaccio). Use when asked to execute integration tests or local publish workflows in this repo.

974 tokens
context cost
the whole folder, loaded on every use
2
files
instructions only
0
copies elsewhere
how many repositories repackaged it
3526
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/openai/openai-agents-python --skill integration-tests

What comes with it

252 bytes besides the instruction
agents/openai.yaml

The instruction itself

9 sections, as written by the author

Integration Tests

Overview

Run integration tests that require a local npm registry by starting pnpm local-npm:start in a background subprocess, executing the build/reset/publish/test pipeline in the main process, then cleanly shutting down the registry process and returning results.

Workflow

Codex execution environment

  • When Codex invokes this skill, run both pnpm local-npm:start and the main pipeline outside the Codex sandbox by default (sandbox_permissions=require_escalated). The integration suites install dependencies inside fixture projects with npm install, bun install, Deno, Wrangler, local emulators, Docker, and other subprocesses; running them inside the Codex sandbox can cause environment-only failures such as npm EPERM writing to ~/.npm/_cacache/tmp or Bun PermissionDenied writing to its tempdir.
  • Use sandboxed execution only when the user explicitly asks for it. If a sandboxed run fails during dependency installation or temp/cache writes, rerun the same exact pipeline outside the sandbox before classifying the failure as repo-induced.
  • Still keep the registry lifecycle local to the run: start Verdaccio before the pipeline and stop it after the pipeline, even when a failure occurs.

1. Start the local registry (subprocess)

  • Start a background process with pnpm local-npm:start and keep its session id so it can be stopped later.
  • Wait until the registry is ready (look for a Verdaccio listen message or the default http://localhost:4873 line). If no explicit ready line appears, wait a few seconds and proceed.
  • If the port is already in use, note that an existing registry may be running and proceed only if it matches the expected local registry; otherwise stop it and restart.

2. Run the main pipeline (main process)

Run this exact sequence in the main process and capture the output:

pnpm i && pnpm build:ci && pnpm local-npm:reset && pnpm local-npm:publish && OPENAI_AGENTS_RUN_STORAGE_MOUNT_INTEGRATION=1 pnpm test:integration
  • Use pnpm build:ci here so the skill validates the same serialized build path that GitHub Actions now uses, while still running the normal prebuild and postbuild lifecycle steps.
  • Run pnpm test:integration with OPENAI_AGENTS_RUN_STORAGE_MOUNT_INTEGRATION=1 so the optional sandbox storage mount coverage runs by default when this skill is used. This enables the local Azurite and MinIO smoke tests without changing the repository's plain pnpm test:integration default.
  • Return the full success/failure outcome and a concise summary of the results.
  • Always capture the stdout/stderr from pnpm test:integration and include it in the final response (trim obvious noise if extremely long) inside a fenced code block.
  • Do not use watch mode.
If pnpm local-npm:publish fails

Troubleshoot using integration-tests/README.md, which lists the canonical recovery steps. If time is short, prioritize the fixes in the order given there and surface the exact error text in your response.

3. Clean up the registry process

  • Send Ctrl+C to the registry subprocess and wait for it to exit.
  • If it does not exit, terminate it by PID and confirm the port is free before finishing.

Output expectations

  • Always include the integration test results in the response.
  • If any step fails, include the failing command, the error output summary, and the next recommended action.

How to use it

Copy the folder

Take openai/integration-tests from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.

Install what it needs

The instructions reference npm. Without those the skill loads but fails at the first command.