Add a Captures tab showing screen captures from the task's agent

internal/agents/codex.go

@@ -169,8 +169,11 @@ func (m *Manager) codexItemCompleted(ctx context.Context, l *live, s *codex.Sess
 		return m.emit(ctx, l, agentActor, "agent.vendor", payload)
 	}
 
-	content, isError := codexToolResult(item)
-	err := m.emit(ctx, l, agentActor, "agent.tool_result", map[string]any{"tool_use_id": item.ID, "is_error": isError, "content": content})
+	content, isError, err := codexToolResult(item)
+	if err != nil {
+		return err
+	}
+	err = m.emit(ctx, l, agentActor, "agent.tool_result", map[string]any{"tool_use_id": item.ID, "is_error": isError, "content": content})
 	if err != nil {
 		return err
 	}
@@ -210,24 +213,51 @@ func imageAlt(prompt string) string {
 	return alt
 }
 
-func codexToolResult(item codex.ThreadItem) (string, bool) {
+func codexToolResult(item codex.ThreadItem) (any, bool, error) {
 	status := codexStatus(item)
 	switch item.Type {
 	case codex.ThreadItemTypeCommandExecution:
-		return item.AggregatedOutput, status != "completed" || item.ExitCode != 0
+		return item.AggregatedOutput, status != "completed" || item.ExitCode != 0, nil
 	case codex.ThreadItemTypeDynamicToolCall:
-		texts := []string{}
+		blocks := []map[string]any{}
 		for _, c := range item.ContentItems {
-			texts = append(texts, c.Text)
+			if c.Type != codex.DynamicToolCallOutputContentItemTypeInputImage {
+				blocks = append(blocks, map[string]any{"type": "text", "text": c.Text})
+				continue
+			}
+			mime, data, ok := strings.Cut(strings.TrimPrefix(c.ImageURL, "data:"), ";base64,")
+			if !ok {
+				return nil, false, fmt.Errorf("dynamic tool image is not a base64 data url")
+			}
+			blocks = append(blocks, imageBlock(mime, data))
 		}
-		return strings.Join(texts, "\n"), !item.Success
+		return blocks, !item.Success, nil
 	case codex.ThreadItemTypeMcpToolCall:
 		if item.Error != nil {
-			return item.Error.Message, true
+			return item.Error.Message, true, nil
 		}
-		return string(item.Result), false
+		res := struct {
+			Content []map[string]any `json:"content"`
+			IsError bool             `json:"isError"`
+		}{}
+		err := json.Unmarshal(item.Result, &res)
+		if err != nil {
+			return nil, false, fmt.Errorf("decode mcp tool result: %w", err)
+		}
+		for i, b := range res.Content {
+			if b["type"] == "image" {
+				mime, _ := b["mimeType"].(string)
+				data, _ := b["data"].(string)
+				res.Content[i] = imageBlock(mime, data)
+			}
+		}
+		return res.Content, res.IsError, nil
 	}
-	return status, status != "completed"
+	return status, status != "completed", nil
+}
+
+func imageBlock(mime, data string) map[string]any {
+	return map[string]any{"type": "image", "source": map[string]any{"type": "base64", "media_type": mime, "data": data}}
 }
 
 func (m *Manager) readCodexRateLimit() {