Protected ReadonlyappProtected ReadonlycustomizationProtected ReadonlyloggerProtected Optional ReadonlymessageProtected ReadonlymigrationProtected ReadonlyopenerProtectedoverrideHow long the prompt customizations have to be stable before overrides are reported. The workspace template locations are configured only after the initial scan of the global ones, and each rescan fires several events; reporting on the first of them would show a notification for a state that is about to change.
ProtectedoverrideSet once the application and the workspace trust are ready, the earliest a report makes sense.
ProtectedoverrideProtectedoverrideProtectedshownProtectedstartupIn-flight run started by onStart or by the initial trust resolution.
Protected ReadonlystorageProtected ReadonlytoProtected ReadonlyworkspaceProtected ReadonlyworkspaceCalled when the application is started. The application shell is not attached yet when this method runs. Should return a promise if it runs asynchronously.
Called when an application is stopped or unloaded.
Note that this is implemented using window.pagehide which doesn't allow any asynchronous code anymore.
I.e. this is the last tick.
Protectedrunwhether to tell the user when there was nothing to migrate. Only the command does: on startup the common case is that nothing is pending, and saying so every time would be noise.
ProtectedrunonDidChangeWorkspaceTrust fires for the initial trust resolution as well as for later
transitions, and that resolution lands after onStart has already started a run and parked it
on the same deferred. Both triggers therefore fire on an ordinary trusted startup. The service
keeps them from migrating twice; this keeps the second one from reporting the result again.
Deliberately an in-flight promise rather than a latch: a workspace that starts untrusted has nothing to migrate, and the run that matters is the one after the user grants trust.
ProtectedscheduleProtectedshowProtectedshow
Runs the
AGENTS.mdmigration on startup and exposes it as a command.The migration writes into the workspace root, a file the user will see in source control, so unlike the silent
customAgents.ymlmigration it always reports what it did. Migration is re-attempted when the workspace becomes trusted; roots added later are covered by the command.