diff --git a/pkg/channels/discord.go b/pkg/channels/discord.go
index 342ddb478..dc92b6256 100644
--- a/pkg/channels/discord.go
+++ b/pkg/channels/discord.go
@@ -3,7 +3,6 @@ package channels
 import (
 	"context"
 	"fmt"
-	"os"
 	"strings"
 	"sync"
 	"time"
@@ -210,19 +209,10 @@ func (c *DiscordChannel) handleMessage(s *discordgo.Session, m *discordgo.Messag
 	content := m.Content
 	content = c.stripBotMention(content)
 	mediaPaths := make([]string, 0, len(m.Attachments))
-	localFiles := make([]string, 0, len(m.Attachments))
-
-	// Ensure temp files are cleaned up when function returns
-	defer func() {
-		for _, file := range localFiles {
-			if err := os.Remove(file); err != nil {
-				logger.DebugCF("discord", "Failed to cleanup temp file", map[string]any{
-					"file":  file,
-					"error": err.Error(),
-				})
-			}
-		}
-	}()
+	// Note: media files in os.TempDir()/picoclaw_media/ are not cleaned up here
+	// because HandleMessage publishes to an async message bus. The consumer
+	// goroutine may still need these files after this function returns.
+	// Temp files are managed by OS temp directory lifecycle.
 
 	for _, attachment := range m.Attachments {
 		isAudio := utils.IsAudioFile(attachment.Filename, attachment.ContentType)
@@ -230,8 +220,7 @@ func (c *DiscordChannel) handleMessage(s *discordgo.Session, m *discordgo.Messag
 		if isAudio {
 			localPath := c.downloadAttachment(attachment.URL, attachment.Filename)
 			if localPath != "" {
-				localFiles = append(localFiles, localPath)
-
+				mediaPaths = append(mediaPaths, localPath)
 				transcribedText := ""
 				if c.transcriber != nil && c.transcriber.IsAvailable() {
 					ctx, cancel := context.WithTimeout(c.getContext(), transcriptionTimeout)
diff --git a/pkg/channels/line.go b/pkg/channels/line.go
index 9f7d2bde0..ec7ac9e72 100644
--- a/pkg/channels/line.go
+++ b/pkg/channels/line.go
@@ -10,7 +10,6 @@ import (
 	"fmt"
 	"io"
 	"net/http"
-	"os"
 	"strings"
 	"sync"
 	"time"
@@ -307,18 +306,10 @@ func (c *LINEChannel) processEvent(event lineEvent) {
 
 	var content string
 	var mediaPaths []string
-	localFiles := []string{}
-
-	defer func() {
-		for _, file := range localFiles {
-			if err := os.Remove(file); err != nil {
-				logger.DebugCF("line", "Failed to cleanup temp file", map[string]interface{}{
-					"file":  file,
-					"error": err.Error(),
-				})
-			}
-		}
-	}()
+	// Note: media files in os.TempDir()/picoclaw_media/ are not cleaned up here
+	// because HandleMessage publishes to an async message bus. The consumer
+	// goroutine may still need these files after this function returns.
+	// Temp files are managed by OS temp directory lifecycle.
 
 	switch msg.Type {
 	case "text":
@@ -330,21 +321,18 @@ func (c *LINEChannel) processEvent(event lineEvent) {
 	case "image":
 		localPath := c.downloadContent(msg.ID, "image.jpg")
 		if localPath != "" {
-			localFiles = append(localFiles, localPath)
 			mediaPaths = append(mediaPaths, localPath)
 			content = "[image]"
 		}
 	case "audio":
 		localPath := c.downloadContent(msg.ID, "audio.m4a")
 		if localPath != "" {
-			localFiles = append(localFiles, localPath)
 			mediaPaths = append(mediaPaths, localPath)
 			content = "[audio]"
 		}
 	case "video":
 		localPath := c.downloadContent(msg.ID, "video.mp4")
 		if localPath != "" {
-			localFiles = append(localFiles, localPath)
 			mediaPaths = append(mediaPaths, localPath)
 			content = "[video]"
 		}
diff --git a/pkg/channels/onebot.go b/pkg/channels/onebot.go
index 06186f783..feeb930af 100644
--- a/pkg/channels/onebot.go
+++ b/pkg/channels/onebot.go
@@ -4,7 +4,6 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
-	"os"
 	"strconv"
 	"strings"
 	"sync"
@@ -571,7 +570,6 @@ type parseMessageResult struct {
 	Text           string
 	IsBotMentioned bool
 	Media          []string
-	LocalFiles     []string
 	ReplyTo        string
 }
 
@@ -603,7 +601,6 @@ func (c *OneBotChannel) parseMessageSegments(raw json.RawMessage, selfID int64)
 	mentioned := false
 	selfIDStr := strconv.FormatInt(selfID, 10)
 	var media []string
-	var localFiles []string
 	var replyTo string
 
 	for _, seg := range segments {
@@ -642,7 +639,6 @@ func (c *OneBotChannel) parseMessageSegments(raw json.RawMessage, selfID int64)
 					})
 					if localPath != "" {
 						media = append(media, localPath)
-						localFiles = append(localFiles, localPath)
 						textParts = append(textParts, fmt.Sprintf("[%s]", segType))
 					}
 				}
@@ -656,7 +652,6 @@ func (c *OneBotChannel) parseMessageSegments(raw json.RawMessage, selfID int64)
 						LoggerPrefix: "onebot",
 					})
 					if localPath != "" {
-						localFiles = append(localFiles, localPath)
 						if c.transcriber != nil && c.transcriber.IsAvailable() {
 							tctx, tcancel := context.WithTimeout(c.ctx, 30*time.Second)
 							result, err := c.transcriber.Transcribe(tctx, localPath)
@@ -703,7 +698,6 @@ func (c *OneBotChannel) parseMessageSegments(raw json.RawMessage, selfID int64)
 		Text:           strings.TrimSpace(strings.Join(textParts, "")),
 		IsBotMentioned: mentioned,
 		Media:          media,
-		LocalFiles:     localFiles,
 		ReplyTo:        replyTo,
 	}
 }
@@ -824,19 +818,10 @@ func (c *OneBotChannel) handleMessage(raw *oneBotRawEvent) {
 		}
 	}
 
-	// Clean up temp files when done
-	if len(parsed.LocalFiles) > 0 {
-		defer func() {
-			for _, f := range parsed.LocalFiles {
-				if err := os.Remove(f); err != nil {
-					logger.DebugCF("onebot", "Failed to remove temp file", map[string]interface{}{
-						"path":  f,
-						"error": err.Error(),
-					})
-				}
-			}
-		}()
-	}
+	// Note: media files in os.TempDir()/picoclaw_media/ are not cleaned up here
+	// because HandleMessage publishes to an async message bus. The consumer
+	// goroutine may still need these files after this function returns.
+	// Temp files are managed by OS temp directory lifecycle.
 
 	if c.isDuplicate(messageID) {
 		logger.DebugCF("onebot", "Duplicate message, skipping", map[string]interface{}{
diff --git a/pkg/channels/slack.go b/pkg/channels/slack.go
index 0060972ed..f7a337195 100644
--- a/pkg/channels/slack.go
+++ b/pkg/channels/slack.go
@@ -3,7 +3,6 @@ package channels
 import (
 	"context"
 	"fmt"
-	"os"
 	"strings"
 	"sync"
 	"time"
@@ -232,19 +231,10 @@ func (c *SlackChannel) handleMessageEvent(ev *slackevents.MessageEvent) {
 	content = c.stripBotMention(content)
 
 	var mediaPaths []string
-	localFiles := []string{} // 跟踪需要清理的本地文件
-
-	// 确保临时文件在函数返回时被清理
-	defer func() {
-		for _, file := range localFiles {
-			if err := os.Remove(file); err != nil {
-				logger.DebugCF("slack", "Failed to cleanup temp file", map[string]interface{}{
-					"file":  file,
-					"error": err.Error(),
-				})
-			}
-		}
-	}()
+	// Note: media files in os.TempDir()/picoclaw_media/ are not cleaned up here
+	// because HandleMessage publishes to an async message bus. The consumer
+	// goroutine may still need these files after this function returns.
+	// Temp files are managed by OS temp directory lifecycle.
 
 	if ev.Message != nil && len(ev.Message.Files) > 0 {
 		for _, file := range ev.Message.Files {
@@ -252,7 +242,6 @@ func (c *SlackChannel) handleMessageEvent(ev *slackevents.MessageEvent) {
 			if localPath == "" {
 				continue
 			}
-			localFiles = append(localFiles, localPath)
 			mediaPaths = append(mediaPaths, localPath)
 
 			if utils.IsAudioFile(file.Name, file.Mimetype) && c.transcriber != nil && c.transcriber.IsAvailable() {
diff --git a/pkg/channels/telegram.go b/pkg/channels/telegram.go
index 20bbf6830..c64fb7126 100644
--- a/pkg/channels/telegram.go
+++ b/pkg/channels/telegram.go
@@ -221,19 +221,10 @@ func (c *TelegramChannel) handleMessage(ctx context.Context, message *telego.Mes
 
 	content := ""
 	mediaPaths := []string{}
-	localFiles := []string{} // 跟踪需要清理的本地文件
-
-	// 确保临时文件在函数返回时被清理
-	defer func() {
-		for _, file := range localFiles {
-			if err := os.Remove(file); err != nil {
-				logger.DebugCF("telegram", "Failed to cleanup temp file", map[string]interface{}{
-					"file":  file,
-					"error": err.Error(),
-				})
-			}
-		}
-	}()
+	// Note: media files in os.TempDir()/picoclaw_media/ are not cleaned up here
+	// because HandleMessage publishes to an async message bus. The consumer
+	// goroutine may still need these files after this function returns.
+	// Temp files are managed by OS temp directory lifecycle.
 
 	if message.Text != "" {
 		content += message.Text
@@ -250,7 +241,6 @@ func (c *TelegramChannel) handleMessage(ctx context.Context, message *telego.Mes
 		photo := message.Photo[len(message.Photo)-1]
 		photoPath := c.downloadPhoto(ctx, photo.FileID)
 		if photoPath != "" {
-			localFiles = append(localFiles, photoPath)
 			mediaPaths = append(mediaPaths, photoPath)
 			if content != "" {
 				content += "\n"
@@ -262,7 +252,6 @@ func (c *TelegramChannel) handleMessage(ctx context.Context, message *telego.Mes
 	if message.Voice != nil {
 		voicePath := c.downloadFile(ctx, message.Voice.FileID, ".ogg")
 		if voicePath != "" {
-			localFiles = append(localFiles, voicePath)
 			mediaPaths = append(mediaPaths, voicePath)
 
 			transcribedText := ""
@@ -297,7 +286,6 @@ func (c *TelegramChannel) handleMessage(ctx context.Context, message *telego.Mes
 	if message.Audio != nil {
 		audioPath := c.downloadFile(ctx, message.Audio.FileID, ".mp3")
 		if audioPath != "" {
-			localFiles = append(localFiles, audioPath)
 			mediaPaths = append(mediaPaths, audioPath)
 			if content != "" {
 				content += "\n"
@@ -309,7 +297,6 @@ func (c *TelegramChannel) handleMessage(ctx context.Context, message *telego.Mes
 	if message.Document != nil {
 		docPath := c.downloadFile(ctx, message.Document.FileID, "")
 		if docPath != "" {
-			localFiles = append(localFiles, docPath)
 			mediaPaths = append(mediaPaths, docPath)
 			if content != "" {
 				content += "\n"