fix(polecat): don't pre-build command for agent override

When --agent is specified, StartSession() pre-built the command via
BuildPolecatStartupCommandWithAgentOverride with an EMPTY prompt.
This set opts.Command, causing session_manager.Start() to skip its
own BuildStartupCommandFromConfig (which includes the beacon prompt).

Result: agent launched without work instructions (no -i flag, no
prompt), sitting idle at the input prompt.

Fix: remove the pre-built command path. Let session_manager.Start()
build the command with both the beacon prompt and agent override
(opts.Agent), which was already being passed through correctly.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: 2 people

internal/cmd/polecat_spawn.go

@@ -358,13 +358,11 @@ func (s *SpawnedPolecatInfo) StartSession() (string, error) {
 		RuntimeConfigDir: claudeConfigDir,
 		Agent:            s.agent,
 	}
-	if s.agent != "" {
-		cmd, err := config.BuildPolecatStartupCommandWithAgentOverride(s.RigName, s.PolecatName, r.Path, "", s.agent)
-		if err != nil {
-			return "", err
-		}
-		startOpts.Command = cmd
-	}
+	// Note: when agent override is specified, we set startOpts.Agent instead of
+	// pre-building the command. This allows session_manager.Start() to build the
+	// command with the beacon prompt included. Previously, pre-building here with
+	// an empty prompt caused the agent to launch without work instructions.
+
 	if err := polecatSessMgr.Start(s.PolecatName, startOpts); err != nil {
 		return "", fmt.Errorf("starting session: %w", err)
 	}