fix(tools): add negative guidance for message tool and disable handoff by default

Message tool prompting now explicitly tells the LLM not to use it for
replying to the user — prevents false activations on phrases like
"gởi lại cho tôi" (send it back to me).

Handoff tool disabled by default since it's rarely used and causes
confusion with spawn. Admins can re-enable via DB/UI if needed.

cmd/gateway_builtin_tools.go

@@ -113,7 +113,7 @@ func builtinToolSeedData() []store.BuiltinToolDef {
 		{Name: "evaluate_loop", DisplayName: "Evaluate Loop", Description: "Run a generate→evaluate→revise loop between two agents for quality-critical output", Category: "delegation", Enabled: true,
 			Requires: []string{"managed_mode", "agent_links"},
 		},
-		{Name: "handoff", DisplayName: "Handoff", Description: "Transfer the conversation to another agent — the user will talk directly to that agent", Category: "delegation", Enabled: true,
+		{Name: "handoff", DisplayName: "Handoff", Description: "Transfer the conversation to another agent — the user will talk directly to that agent", Category: "delegation", Enabled: false,
 			Requires: []string{"managed_mode", "agent_links"},
 		},
 

internal/agent/systemprompt.go

@@ -70,7 +70,7 @@ var coreToolSummaries = map[string]string{
 	"browser":          "Browse web pages interactively",
 	"tts":              "Convert text to speech audio",
 	"edit":             "Edit a file by replacing exact text matches",
-	"message":          "Send a message to a channel (Telegram, Discord, etc.)",
+	"message":          "Send a PROACTIVE message to another channel/chat — do NOT use this to reply to the user, just respond directly",
 	"sessions_list":    "List sessions for this agent",
 	"session_status":   "Show session status (model, tokens, compaction count)",
 	"sessions_history": "Fetch message history for a session",