A checkout scenario fails only on the CI machine. I can open the trace with sdods trace --last and it is clear enough what happened, but the developer who has to fix it does not have SDODS installed and is not going to install it to look at one failure.
What do I actually send them? The whole run directory is 300 MB and most of it is step screenshots they do not need.
Accepted answer
Send the one trace.zip. Find it first:
sdods trace --last --list--list prints the trace files instead of opening them, and with --json you get { source, traces } if you are scripting it. Then hand over the single zip — they drop it on trace.playwright.dev, which runs entirely in their browser and uploads nothing, so it is fine for a trace of an internal environment.
Worth knowing: traces are recorded on the first retry and for failures, so a fully green run has none and sdods trace exits 2 with nothing to open. If the developer says "can you get me a trace of it passing", you have to make it fail or run with retries first.
If they would rather not open a viewer at all, two files usually do it: the dashboard index.html, which is self-contained, and the failure screenshot at <runDir>/<project-slug>/<fingerprint>/r<attempt>/failure.png. That is enough for most "what did the page look like" questions and it is two attachments instead of 300 MB.
Note the attempt number — a failure that was retried is under r1, not r0, and people send the wrong one constantly.
Have a different question? Ask it.