From 5d22b29838f8ce665fa74527c18b13e26aa0fd92 Mon Sep 17 00:00:00 2001
From: LifeJiggy <Bloomtonjovish@gmail.com>
Date: Tue, 26 May 2026 01:37:33 +0100
Subject: [PATCH] fix(shell): preview Unicode truncation uses byte length
 instead of char length

The preview function used text.length (UTF-16 code units) to check
if output exceeds the metadata limit, but the actual truncation
should be based on byte length for correct handling of multi-byte
characters (CJK, emoji, etc.). Use Buffer.byteLength for the check
and ensure truncated output does not split a multi-byte character.
---
 packages/opencode/src/tool/shell.ts       |  7 +++++--
 packages/opencode/test/tool/shell.test.ts | 17 +++++++++++++++++
 2 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/packages/opencode/src/tool/shell.ts b/packages/opencode/src/tool/shell.ts
index b6a95b5c0970..a412f760fb3d 100644
--- a/packages/opencode/src/tool/shell.ts
+++ b/packages/opencode/src/tool/shell.ts
@@ -221,8 +221,11 @@ function pathArgs(list: Part[], ps: boolean, cmd = false) {
 }
 
 function preview(text: string) {
-  if (text.length <= MAX_METADATA_LENGTH) return text
-  return "...\n\n" + text.slice(-MAX_METADATA_LENGTH)
+  if (Buffer.byteLength(text, "utf-8") <= MAX_METADATA_LENGTH) return text
+  let buf = Buffer.from(text, "utf-8")
+  buf = buf.subarray(-MAX_METADATA_LENGTH)
+  while (buf.length > 0 && (buf[0] & 0xc0) === 0x80) buf = buf.subarray(1)
+  return "...\n\n" + buf.toString("utf-8")
 }
 
 function tail(text: string, maxLines: number, maxBytes: number) {
diff --git a/packages/opencode/test/tool/shell.test.ts b/packages/opencode/test/tool/shell.test.ts
index ddaa5c2ec7b1..11bf2397b2d5 100644
--- a/packages/opencode/test/tool/shell.test.ts
+++ b/packages/opencode/test/tool/shell.test.ts
@@ -1227,4 +1227,21 @@ describe("tool.shell truncation", () => {
       }),
     ),
   )
+
+  it.live("preview keeps valid UTF-8 for multi-byte output", () =>
+    runIn(
+      projectRoot,
+      Effect.gen(function* () {
+        const n = 12000
+        const result = yield* run({
+          command: `${bin} -e ${evalarg(`process.stdout.write('\\u4e2d'.repeat(${n}))`)} ${n}`,
+          description: "Generate Unicode output",
+        })
+        const meta = result.metadata as { output?: string }
+        if (meta.output) {
+          expect(meta.output.includes("\uFFFD")).toBe(false)
+        }
+      }),
+    ),
+  )
 })