Hi team,
I am reporting a regression that appeared after updating to the latest version (v1.23+).
The Issue:
The terminal session used by the AI agent is no longer interactive or accessible for the user.
Technical Detail:
I have the “Enable Shell Integration” option enabled in the Agent settings. Despite this being enabled (which should use the IDE’s shell integration to report terminal execution), the terminal remains non-interactive.
Why this matters:
Previously, we could see and interact with the agent’s terminal. This was critical for entering sudo passwords or other sensitive information directly into the shell.
Impact:
Now, if an agent runs a command requiring sudo, the process hangs or fails because there is no way for the user to provide the password. We are forced to terminate the agent’s task and run the command manually in a separate terminal, which significantly disrupts the agentic development workflow.
Request:
Please restore terminal interactivity or provide a secure mechanism for users to provide interactive input (like passwords) to the agent’s terminal session while Shell Integration is active.
Thank you!