bug(cli): sandbox create with duplicate name returns raw database error instead of user-friendly message

[areporeporepo]

agent diagnostic

• loaded openshell-cli skill — checked sandbox create --help for --force/--replace/--overwrite flags; none exist
• loaded debug-openshell-cluster skill — cluster is healthy (openshell status and openshell doctor check both pass)
• confirmed workaround: sandbox delete <name> then sandbox create --name <name> works, but the error message gives no hint to do this
• agent cannot resolve: the error originates in the grpc server layer — sqlite unique constraint error (code 2067) is surfaced as a raw internal grpc status instead of being caught and mapped to alreadyexists with a human-readable message. this requires a server-side code change in the persist layer or grpc handler.

description

actual behavior: creating a sandbox with a name that already exists returns a raw internal error exposing sqlite internals and grpc metadata:

error: × status: Internal, message: "persist sandbox failed: execution error:
  database error: error returned from database: (code: 2067) UNIQUE
  constraint failed: objects.object_type, objects.name", details: [],
  metadata: MetadataMap { headers: {"content-type": "application/grpc",
  "date": "Mon, 30 Mar 2026 23:51:20 GMT"} }

expected behavior: a clear, actionable error like:

error: sandbox 'dupe-test' already exists

hint: delete it first with: openshell sandbox delete dupe-test
      or use a different name

reproduction steps

1. openshell sandbox create --name dupe-test --no-tty -- sleep 120 &
2. wait for sandbox to reach ready phase (openshell sandbox list)
3. openshell sandbox create --name dupe-test --no-tty -- echo "second"
4. observe raw database error instead of "already exists" message

environment

• os: macos 26.3.1 (darwin, apple silicon)
• docker: docker desktop 29.2.1
• openshell: v0.0.6

logs

$ openshell sandbox create --name dupe-test --no-tty -- echo "second"
Error:   × status: Internal, message: "persist sandbox failed: execution error:
  │ database error: error returned from database: (code: 2067) UNIQUE
  │ constraint failed: objects.object_type, objects.name", details: [],
  │ metadata: MetadataMap { headers: {"content-type": "application/grpc",
  │ "date": "Mon, 30 Mar 2026 23:51:20 GMT"} }

[areporeporepo]

suggested fix

the server already handles the k8s duplicate case correctly at crates/openshell-server/src/grpc.rs:273:

return Err(Status::already_exists("sandbox already exists"));

the store persist at line 258 just needs the same pattern. the codebase already checks for UNIQUE constraint errors elsewhere (line 2620-2621):

if msg.contains("UNIQUE") || msg.contains("unique") || msg.contains("duplicate") {

suggested change at line 256-258:

// before (current)
.map_err(|e| Status::internal(format!("persist sandbox failed: {e}")))?;

// after (fix)
.map_err(|e| {
    let msg = e.to_string();
    if msg.contains("UNIQUE") || msg.contains("unique") {
        Status::already_exists(format!("sandbox '{}' already exists", name))
    } else {
        Status::internal(format!("persist sandbox failed: {e}"))
    }
})?;

same fix applies to the second put_message call at line 612-614.

the cli side could also validate client-side before the grpc call using sandbox list, but the server-side fix is the right layer since other clients (tui, api) would hit the same error.