diff --git a/docs/release-control/v6/internal/subsystems/ai-runtime.md b/docs/release-control/v6/internal/subsystems/ai-runtime.md
index 77bc2fe5b..e263e917a 100644
--- a/docs/release-control/v6/internal/subsystems/ai-runtime.md
+++ b/docs/release-control/v6/internal/subsystems/ai-runtime.md
@@ -743,6 +743,22 @@ runtime cost control, and shared AI transport surfaces.
    arrives; it must not flash as run-on prose such as
    `I'llcheckthedevicenodes...` while the actual governed tool row is still
    being assembled.
+   A pre-tool assistant preamble such as "I'll check that" does not satisfy the
+   final-answer contract after Pulse executes tools. The final-response guard
+   must require non-empty assistant text after the latest user/tool-result
+   anchor; otherwise it must make the bounded no-tools summary call or emit the
+   deterministic fallback summary. This adapts the referenced OpenCode source
+   at fetched `origin/dev` commit
+   `06d7840d1d42c9815d2d2e45e7fa4090ca4e3577`: `packages/opencode/src/session/llm/ai-sdk.ts`
+   maps tool input/call/result stream events into typed tool events (lines
+   190-246), `packages/opencode/src/session/message-v2.ts` serializes completed
+   tool results and reasoning as ordered provider message parts (lines
+   340-390), and `packages/ui/src/components/message-part.tsx` only renders
+   non-empty typed text/reasoning/tool parts through `renderable` and
+   `AssistantParts` (lines 607-728). Pulse adapts that ordered-part invariant
+   inside `internal/ai/chat/agentic_final.go` because Pulse's transcript is
+   persisted as messages plus tool-result anchors rather than OpenCode's
+   per-part store.
    Streamed provider startup and mid-stream progress must be bounded by the
    configured Assistant request timeout, the OpenAI-compatible SSE
    response-header guard, and the per-chunk SSE body-read guard adapted from
diff --git a/internal/ai/chat/agentic_additional_test.go b/internal/ai/chat/agentic_additional_test.go
index 3c5dd423e..1ef9f4b25 100644
--- a/internal/ai/chat/agentic_additional_test.go
+++ b/internal/ai/chat/agentic_additional_test.go
@@ -187,6 +187,112 @@ func TestEnsureFinalTextResponse(t *testing.T) {
 	}
 }
 
+func TestEnsureFinalTextResponseRequiresAssistantTextAfterLatestToolResult(t *testing.T) {
+	provider := &stubStreamingProvider{}
+	loop := &AgenticLoop{provider: provider, baseSystemPrompt: "prompt"}
+
+	var emitted strings.Builder
+	result := loop.ensureFinalTextResponse(
+		context.Background(),
+		"session-pre-tool-preamble",
+		[]Message{
+			{Role: "user", Content: "how many devices are in this?"},
+			{
+				Role:    "assistant",
+				Content: "I'll check the device nodes.",
+				ToolCalls: []ToolCall{
+					{ID: "call-1", Name: "pulse_read", Input: map[string]interface{}{"command": "ls /dev | wc -l"}},
+				},
+			},
+			{
+				Role:       "user",
+				ToolResult: &ToolResult{ToolUseID: "pulse_read_0", Content: "4358", IsError: false},
+			},
+			{Role: "assistant", Content: ""},
+		},
+		[]providers.Message{
+			{Role: "user", Content: "how many devices are in this?"},
+			{
+				Role:    "assistant",
+				Content: "I'll check the device nodes.",
+				ToolCalls: []providers.ToolCall{
+					{ID: "call-1", Name: "pulse_read", Input: map[string]interface{}{"command": "ls /dev | wc -l"}},
+				},
+			},
+			{Role: "tool", ToolResult: &providers.ToolResult{ToolUseID: "call-1", Content: "4358"}},
+			{Role: "assistant", Content: ""},
+		},
+		func(event StreamEvent) {
+			if event.Type != "content" {
+				return
+			}
+			var data ContentData
+			if err := json.Unmarshal(event.Data, &data); err == nil {
+				emitted.WriteString(data.Text)
+			}
+		},
+	)
+
+	if len(result) != 5 {
+		t.Fatalf("expected final summary to be appended after tool result, got %d messages", len(result))
+	}
+	if result[len(result)-1].Content != "summary" {
+		t.Fatalf("expected provider summary to be appended, got %q", result[len(result)-1].Content)
+	}
+	if emitted.String() != "summary" {
+		t.Fatalf("expected summary to be streamed, got %q", emitted.String())
+	}
+}
+
+func TestHasFinalAssistantText(t *testing.T) {
+	tests := []struct {
+		name     string
+		messages []Message
+		want     bool
+	}{
+		{
+			name:     "assistant text without tool results is final",
+			messages: []Message{{Role: "assistant", Content: "done"}},
+			want:     true,
+		},
+		{
+			name: "assistant text before latest user prompt is not final",
+			messages: []Message{
+				{Role: "assistant", Content: "previous answer"},
+				{Role: "user", Content: "new question"},
+			},
+			want: false,
+		},
+		{
+			name: "pre tool text before latest tool result is not final",
+			messages: []Message{
+				{Role: "user", Content: "check it"},
+				{Role: "assistant", Content: "I'll check", ToolCalls: []ToolCall{{Name: "pulse_query"}}},
+				{Role: "user", ToolResult: &ToolResult{ToolUseID: "pulse_query_0", Content: "ok"}},
+			},
+			want: false,
+		},
+		{
+			name: "assistant text after latest tool result is final",
+			messages: []Message{
+				{Role: "user", Content: "check it"},
+				{Role: "assistant", Content: "I'll check", ToolCalls: []ToolCall{{Name: "pulse_query"}}},
+				{Role: "user", ToolResult: &ToolResult{ToolUseID: "pulse_query_0", Content: "ok"}},
+				{Role: "assistant", Content: "The result is ok."},
+			},
+			want: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if got := hasFinalAssistantText(tt.messages); got != tt.want {
+				t.Fatalf("hasFinalAssistantText() = %v, want %v", got, tt.want)
+			}
+		})
+	}
+}
+
 func TestEnsureFinalTextResponseAppliesRequestSanitizer(t *testing.T) {
 	provider := &stubStreamingProvider{}
 	loop := &AgenticLoop{provider: provider, baseSystemPrompt: "raw-host"}
diff --git a/internal/ai/chat/agentic_final.go b/internal/ai/chat/agentic_final.go
index bd13d82e6..15484ac85 100644
--- a/internal/ai/chat/agentic_final.go
+++ b/internal/ai/chat/agentic_final.go
@@ -13,9 +13,9 @@ import (
 	"github.com/rs/zerolog/log"
 )
 
-// ensureFinalTextResponse checks if the result messages contain any assistant text.
-// If not, it makes one last LLM call without tools so the model can summarize
-// its findings.
+// ensureFinalTextResponse checks if the result messages contain assistant text
+// after the latest user/tool-result anchor. If not, it makes one last LLM call
+// without tools so the model can summarize its findings.
 // This prevents the loop from exiting silently after making tool calls without answering.
 //
 // cost-recording-exempt: any tokens this final summary turn consumes
@@ -30,11 +30,8 @@ func (a *AgenticLoop) ensureFinalTextResponse(
 	providerMessages []providers.Message,
 	callback StreamCallback,
 ) []Message {
-	// Check if any assistant message has text content
-	for i := len(resultMessages) - 1; i >= 0; i-- {
-		if resultMessages[i].Role == "assistant" && strings.TrimSpace(resultMessages[i].Content) != "" {
-			return resultMessages // Already has text — nothing to do
-		}
+	if hasFinalAssistantText(resultMessages) {
+		return resultMessages
 	}
 
 	// No text content from the model. Make a final text-only call.
@@ -147,6 +144,26 @@ func (a *AgenticLoop) ensureFinalTextResponse(
 	return resultMessages
 }
 
+func hasFinalAssistantText(resultMessages []Message) bool {
+	anchor := -1
+	for i, msg := range resultMessages {
+		if msg.ToolResult != nil {
+			anchor = i
+			continue
+		}
+		if msg.Role == "user" && strings.TrimSpace(msg.Content) != "" {
+			anchor = i
+		}
+	}
+
+	for i := len(resultMessages) - 1; i > anchor; i-- {
+		if resultMessages[i].Role == "assistant" && strings.TrimSpace(resultMessages[i].Content) != "" {
+			return true
+		}
+	}
+	return false
+}
+
 func buildAutomaticFallbackSummary(resultMessages []Message) string {
 	successCount := 0
 	errorCount := 0