Some problems with Diagnosis

1. Capture:
When Capturing in pods like kubeskoop's "pod controller" , meet this error: "failed to run command: pod not found on nettop cache".
When Capturing in pods like kubeskoop's "pod agent", it meet this error: "failed to upload task result", and "pod controller" seems to crash and reload.

2. Connectivity Diagnosis:
How could I submit args like "--network-plugin", when auto-detection failed on web console?

Start by reproducing the two Capture failures in the pod controller and pod agent, then inspect how the web console handles Connectivity Diagnosis when auto-detection fails. Done should include successful capture without the cache, upload, or controller reload errors, plus a documented or working way to submit the --network-plugin argument.

Written by the indexing model from the issue text.