Debug an npx CLI Failure Without a Screenshare

npx can fetch a newer dist-tag than the quickstart you wrote, or a cached older one. If the error message does not print the resolved version, you will debug a release the developer is not running.

Print name, version, and node version on every failure path. Pin the quickstart to npx package@version when a breaking change is in flight. 'Latest' is how Thursday's docs break Friday's eval.

A screenshare shows one laptop and burns an hour. The probe shows the command, the resolved version, the exit status, and what they were trying to do. That is the reproduction.

Strip tokens from argv before you store it. Flags like --token and URLs with query secrets are the usual leaks. Store the flag names, not the values, unless the value is an enum such as --env staging.

If the CLI cannot reach your API, your tracing backend stays dark. The interesting part happened on their machine: DNS, a corporate proxy, Docker not running. Only a probe sent from the CLI, or a local record they can forward, contains it.

Timeouts on the probe itself must not change the exit code. Their failure remains the failure.

When the draft is clear, answer with the exact command that works on that version. If many drafts share a version, yank or fix the release. The screenshare was never the scarce resource. The reproduction was.