From 97a4c64cc807b2513d3e9f936edcb0f2f882f34a Mon Sep 17 00:00:00 2001
From: ZF <130743669+feizhuzheng@users.noreply.github.com>
Date: Wed, 1 Jul 2026 16:45:26 +0800
Subject: [PATCH] fix(harness): truncate text on rune boundary to keep UTF-8
 valid (#16511)

### Summary

`truncateText` in the `reduction` and `summarization` middlewares
truncates with `s[:maxLen]`, which slices by byte. When `maxLen` lands
inside a multi-byte character (common with CJK or other non-ASCII
content flowing through the agent), the string is cut mid-rune and the
tail byte(s) become invalid UTF-8. That broken text then goes into the
reduced context / summary prompt.

`TruncateToolResult` in the same `reduction` package already avoids this
by slicing on a rune boundary and even notes it in a comment. This PR
makes the two `truncateText` helpers do the same, so they stay
consistent with the existing helper.

Both functions keep their existing output shape (summarization still
appends `...`). Added a small unit test in each package covering ASCII
truncation and a CJK string, asserting the result stays valid UTF-8.
---
 .../core/middlewares/reduction/reduction.go   |  8 +++++-
 .../reduction/truncate_text_test.go           | 25 ++++++++++++++++++
 .../summarization/summarization.go            |  8 +++++-
 .../summarization/truncate_text_test.go       | 26 +++++++++++++++++++
 4 files changed, 65 insertions(+), 2 deletions(-)
 create mode 100644 internal/harness/core/middlewares/reduction/truncate_text_test.go
 create mode 100644 internal/harness/core/middlewares/summarization/truncate_text_test.go

diff --git a/internal/harness/core/middlewares/reduction/reduction.go b/internal/harness/core/middlewares/reduction/reduction.go
index b2ee8a9fa..073ad9d6c 100644
--- a/internal/harness/core/middlewares/reduction/reduction.go
+++ b/internal/harness/core/middlewares/reduction/reduction.go
@@ -165,5 +165,11 @@ func truncateText(s string, maxLen int) string {
 	if len(s) <= maxLen {
 		return s
 	}
-	return s[:maxLen]
+	// Slice on a rune boundary; [:maxLen] indexes bytes and can split a
+	// multi-byte UTF-8 character, same as TruncateToolResult guards against.
+	runes := []rune(s)
+	if maxLen > len(runes) {
+		return s
+	}
+	return string(runes[:maxLen])
 }
diff --git a/internal/harness/core/middlewares/reduction/truncate_text_test.go b/internal/harness/core/middlewares/reduction/truncate_text_test.go
new file mode 100644
index 000000000..d9a8f0e2b
--- /dev/null
+++ b/internal/harness/core/middlewares/reduction/truncate_text_test.go
@@ -0,0 +1,25 @@
+package reduction
+
+import (
+	"testing"
+	"unicode/utf8"
+)
+
+func TestTruncateText(t *testing.T) {
+	if got := truncateText("hello", 10); got != "hello" {
+		t.Fatalf("short string changed: %q", got)
+	}
+	if got := truncateText("hello world", 5); got != "hello" {
+		t.Fatalf("ascii truncation: got %q want %q", got, "hello")
+	}
+
+	// A byte slice at maxLen would land in the middle of a multi-byte rune
+	// and yield invalid UTF-8. Truncating on a rune boundary must not.
+	got := truncateText("日本語のテキスト", 3)
+	if !utf8.ValidString(got) {
+		t.Fatalf("result is not valid UTF-8: %q", got)
+	}
+	if got != "日本語" {
+		t.Fatalf("got %q want %q", got, "日本語")
+	}
+}
diff --git a/internal/harness/core/middlewares/summarization/summarization.go b/internal/harness/core/middlewares/summarization/summarization.go
index dc81dbce3..a6afe386a 100644
--- a/internal/harness/core/middlewares/summarization/summarization.go
+++ b/internal/harness/core/middlewares/summarization/summarization.go
@@ -275,7 +275,13 @@ func truncateText(s string, maxLen int) string {
 	if len(s) <= maxLen {
 		return s
 	}
-	return s[:maxLen] + "..."
+	// Slice on a rune boundary; [:maxLen] indexes bytes and can split a
+	// multi-byte UTF-8 character, producing invalid output in the summary.
+	runes := []rune(s)
+	if maxLen > len(runes) {
+		return s
+	}
+	return string(runes[:maxLen]) + "..."
 }
 
 func buildSummaryPrompt[M core.MessageType](content string) M {
diff --git a/internal/harness/core/middlewares/summarization/truncate_text_test.go b/internal/harness/core/middlewares/summarization/truncate_text_test.go
new file mode 100644
index 000000000..1dea7b15c
--- /dev/null
+++ b/internal/harness/core/middlewares/summarization/truncate_text_test.go
@@ -0,0 +1,26 @@
+package summarization
+
+import (
+	"strings"
+	"testing"
+	"unicode/utf8"
+)
+
+func TestTruncateText(t *testing.T) {
+	if got := truncateText("hello", 10); got != "hello" {
+		t.Fatalf("short string changed: %q", got)
+	}
+	if got := truncateText("hello world", 5); got != "hello..." {
+		t.Fatalf("ascii truncation: got %q want %q", got, "hello...")
+	}
+
+	// A byte slice at maxLen would split a multi-byte rune and produce
+	// invalid UTF-8; truncating on a rune boundary must not.
+	got := truncateText("日本語のテキスト", 3)
+	if !utf8.ValidString(strings.TrimSuffix(got, "...")) {
+		t.Fatalf("result is not valid UTF-8: %q", got)
+	}
+	if got != "日本語..." {
+		t.Fatalf("got %q want %q", got, "日本語...")
+	}
+}