Start app during init to verify instrumentation

[sentry-junior]

After sentry init instruments a project, nothing reaches Sentry unless the app is actually running and exercised. Users are left to know they should start the app and try things out; the CLI does not detect a local runtime or drive a verification path.

Agent plugins already do a stronger version of this: figure out how to start the app (or ask the user), execute test code, and send a test error so setup is proven end-to-end.

Current behavior

• sentry init configures the SDK and finishes without checking whether the app is running locally
• no start/run step and no automatic test error from a live app process
• success can look complete while zero events were sent

Gap

• Local instrumentation only produces signal when the app runs; without that, users get a false sense that setup is done
• Naive always-run behavior is risky on large/hard projects (e.g. monorepos like sentry), where agent-driven start attempts can burn excessive tokens

Related: #1381 (init success output does not prove setup worked)

Proposal from discussion

• Try to start/run the project after init, similar to the plugin flow, and send a test error when feasible
• Bail out early (or ask the user) when the project looks difficult/expensive to run rather than forcing a full start attempt

Raised by Dhrumil Parekh; constrained by Miguel Betegon from prior run attempts on large repos.

Requested by Miguel Betegon.

--

View Junior Session [Sentry]