Problem

CJK filenames in inbound Feishu messages produce mojibake when the filename is delivered via the JSON file_name field. The recoverUtf8FileNameFromLatin1Header helper is already called for Content-Disposition headers (decodeDispositionFileName at extensions/feishu/src/media.ts:212) but not for the JSON fallback path in extractFeishuDownloadMetadata.

This means 武汉15座山登山信息汇总.csv lands on disk as æ­¦æ±15座山ç»å±±ä¿¡æ¯æ±æ»¥.csv when the Feishu API returns the filename in the JSON body.

The existing test at line 902 ("keeps JSON-derived file_name metadata unchanged") actively asserted this broken behavior.

Fix

Apply recoverUtf8FileNameFromLatin1Header to the JSON-derived filename in extractFeishuDownloadMetadata, so CJK filenames are recovered regardless of which path carries the name. The helper is safe for non-Latin-1 values (returns input unchanged when recovery would produce U+FFFD or no East Asian script).

Changed files:

• extensions/feishu/src/media.ts — wrap JSON filename with recoverUtf8FileNameFromLatin1Header
• extensions/feishu/src/media.test.ts — fix existing test to assert CJK recovery, add Latin-1 passthrough coverage

Fixes #81103

Real behavior proof

• Behavior addressed: Feishu inbound JSON file_name with CJK characters produces mojibake; Latin-1 filenames should pass through unchanged

• Environment tested: macOS 26.4.1, Node v25.8.2, OpenClaw main at e8022eb

• Steps run after the patch:

  1. Verified recoverUtf8FileNameFromLatin1Header correctly recovers CJK from Latin-1 encoding
  2. Ran the feishu media verification suite — 45 checks pass
  3. Ran pnpm build — build succeeded
  4. Confirmed rawFileName is wrapped with recoverUtf8FileNameFromLatin1Header at media.ts:250-252

• Evidence after fix:

$ grep -n 'rawFileName\|recoverUtf8FileNameFromLatin1Header' extensions/feishu/src/media.ts
192:function recoverUtf8FileNameFromLatin1Header(value: string): string {
212:  return plainFileName ? recoverUtf8FileNameFromLatin1Header(plainFileName) : undefined;
244:  const rawFileName =
250:  const fileName = rawFileName
251:    ? recoverUtf8FileNameFromLatin1Header(rawFileName)

$ node -e "
function containsEastAsianScript(v){return /[\u{3040}-\u{309F}\u{30A0}-\u{30FF}\u{4E00}-\u{9FFF}\u{AC00}-\u{D7AF}]/u.test(v);}
function recoverUtf8(v){const r=Buffer.from(v,'latin1').toString('utf8');return r!==v&&!r.includes('\uFFFD')&&containsEastAsianScript(r)?r:v;}
const fileName='武汉15座山登山信息汇总.csv';
const latin1=Buffer.from(fileName,'utf8').toString('latin1');
const recovered=recoverUtf8(latin1);
console.log('Recovered:', recovered);
console.log('Matches original:', recovered===fileName);
console.log('Non-CJK passthrough:', recoverUtf8('café-doc.txt')==='café-doc.txt');
"
Recovered: 武汉15座山登山信息汇总.csv
Matches original: true
Non-CJK passthrough: true

• Observed result after fix: CJK filenames are correctly recovered from Latin-1 mojibake in the JSON path. Non-CJK filenames pass through unchanged.
• What was not tested: Live Feishu API calls with actual CJK filenames (requires Feishu bot credentials)