This is a comment left during a code review.
Path: extensions/feishu/src/media.ts
Line: 459-473

Comment:
**`probeBufferDurationMs` can throw, aborting the entire media send**

Unlike `probeMediaDurationMs` which wraps all operations in `try/catch` and returns `undefined` on failure, `probeBufferDurationMs` has no error handling around `writeFile`. If the temp file write fails (e.g. disk full, permissions issue), the exception propagates all the way up through `sendMediaFeishu`, and no message is delivered at all — even though duration metadata is intended to be best-effort.

This is inconsistent with how `transcodeAudioToFeishuOpus` (returns `null` on failure) and `probeMediaDurationMs` (returns `undefined` on failure) behave. The call site in `sendMediaFeishu` doesn't add a `.catch()` either.

```suggestion
async function probeBufferDurationMs(params: {
  buffer: Buffer;
  fileName: string;
}): Promise<number | undefined> {
  const ext = path.extname(params.fileName).toLowerCase() || ".bin";
  const tmpDir = await fs.promises.mkdtemp(path.join(os.tmpdir(), "openclaw-feishu-probe-"));
  const filePath = path.join(tmpDir, `probe${ext}`);

  try {
    await fs.promises.writeFile(filePath, params.buffer);
    return await probeMediaDurationMs(filePath);
  } catch {
    return undefined;
  } finally {
    await fs.promises.rm(tmpDir, { recursive: true, force: true }).catch(() => undefined);
  }
}
```

How can I resolve this? If you propose a fix, please make it concise.

This is a comment left during a code review.
Path: extensions/feishu/src/media.ts
Line: 431-435

Comment:
**Transcode failures are silently swallowed with no log**

When `ffmpeg` fails (binary not installed, malformed input, out of disk space, etc.) the `catch` block returns `null` and the code falls back to uploading the original file as type `"stream"` with `msg_type: "file"`. From the user's perspective the voice message silently becomes a file download, and from an operator's perspective there is no log entry to diagnose why audio delivery degraded.

Consider at minimum emitting a warning through the runtime logger so that transcode failures are observable in production:

```typescript
  } catch (err) {
    // Returning null here causes sendMediaFeishu to fall back to the original file as "stream"
    // Log the failure so it is visible in production diagnostics
    params.runtime?.error?.(`feishu: audio transcode failed for "${params.fileName}": ${String(err)}`);
    return null;
  }
```

(A runtime reference would need to be threaded through, or at least a `console.warn` could serve as a minimal indicator.)

How can I resolve this? If you propose a fix, please make it concise.

This is a comment left during a code review.
Path: src/infra/outbound/payloads.ts
Line: 40-46

Comment:
**Duplicate `stripInternalTtsTags` implementation**

The exact same function body is copy-pasted into `extensions/feishu/src/reply-dispatcher.ts` (lines 24–30). If the tag format ever changes, both sites need to be updated in sync.

Consider extracting this to a shared utility in `openclaw/plugin-sdk` or a common infra helper so there is a single source of truth.

How can I resolve this? If you propose a fix, please make it concise.