-
Notifications
You must be signed in to change notification settings - Fork 394
Handling password protected commands #1551
Copy link
Copy link
Labels
area/securityAuthentication, authorization, secrets, vulnerabilitiesAuthentication, authorization, secrets, vulnerabilitiesarea/toolsFor features/issues/fixes related to the usage of built-in and MCP toolsFor features/issues/fixes related to the usage of built-in and MCP toolskind/featPR adds a new feature (maps to feat: commit prefix)PR adds a new feature (maps to feat: commit prefix)
Metadata
Metadata
Assignees
Labels
area/securityAuthentication, authorization, secrets, vulnerabilitiesAuthentication, authorization, secrets, vulnerabilitiesarea/toolsFor features/issues/fixes related to the usage of built-in and MCP toolsFor features/issues/fixes related to the usage of built-in and MCP toolskind/featPR adds a new feature (maps to feat: commit prefix)PR adds a new feature (maps to feat: commit prefix)
Type
Fields
Give feedbackNo fields configured for Enhancement.
Sometimes the default agents will issue commands that require interactive user input, e.g. sudo somecommand.
After a few retries, the agent usually detects the command is hanging and falls back to providing manual instructions for the user to continue.
What would be an elegant way to support a one-time privilege escalation flow (e.g., prompt once for sudo approval/password, then continue non-interactively)?