After clicking stop, the task still outputs—how to confirm if stopping was successful?
Stop requests may fail or encounter runtime switching; do not assume button changes mean all operations have been revoked.
Stopping involves both runtime and display
The page stop operation attempts to cancel Agent execution and stop the current response stream; these two steps are not exactly the same. The interface activity indicator may change first; if canceling execution fails, a prompt will show that stopping failed and to please retry. Therefore, output no longer scrolling does not equal server confirmation that the task has stopped, nor can closing the page replace checking the stop result.
Continue processing based on the latest runtime status
If the runtime has switched, the interface may prompt to reconnect to the current runtime instead of canceling the task corresponding to the old state. Wait for the latest status and output to appear before deciding whether to stop again. When the network is unavailable, first restore the connection; for local tasks, also verify the actual status of the target machine. Do not immediately submit the same task when the stop result is unclear, as this may cause two executions with side effects.
Stopping does not undo changes already made
Files already written, software installed, or network requests already sent will not automatically roll back because of the stop button. After confirming stop, check the results and decide how to recover or continue. For CLI clients that need to exit, press Ctrl+C in their running terminal, but also verify whether cleanup succeeded. For important files, first perform read-only checks before executing necessary repairs; feedback only provides desensitized status and errors, without sharing full connection credentials.