diff --git a/internal/helpers/doc.go b/internal/helpers/doc.go
index 29fba2e5..4c68b914 100644
--- a/internal/helpers/doc.go
+++ b/internal/helpers/doc.go
@@ -362,7 +362,7 @@ func newDocCreateCommand(runner executor.Runner) *cobra.Command {
 						content = stripped
 					}
 					if content != "" {
-						params["markdown"] = content
+						params["markdown"] = stripDocInputUnsafe(content)
 					}
 				}
 			} else if format, err := docContentFormat(cmd, "markdown", "jsonml"); err != nil {
@@ -473,7 +473,7 @@ func newDocUpdateCommand(runner executor.Runner) *cobra.Command {
 				if sniffJsonMLLike(content) {
 					fmt.Fprintln(cmd.ErrOrStderr(), `warning: 输入内容看起来是 JSONML 结构；若要按 JSONML 解析，请加 --content-format jsonml，否则将按 markdown 解析。`)
 				}
-				params["markdown"] = content
+				params["markdown"] = stripDocInputUnsafe(content)
 				addDocIntParam(cmd, params, "index", "index")
 			}
 			return runDocTool(cmd, runner, "doc", "update_document", params)
diff --git a/internal/helpers/doc_jsonml.go b/internal/helpers/doc_jsonml.go
index b3b8c92b..42fc0856 100644
--- a/internal/helpers/doc_jsonml.go
+++ b/internal/helpers/doc_jsonml.go
@@ -26,6 +26,7 @@ import (
 var docDangerousUnicode = [...]rune{
 	0x200B,
 	0x200C,
+	0x200D,
 	0x200E,
 	0x200F,
 	0x202A,
@@ -33,6 +34,8 @@ var docDangerousUnicode = [...]rune{
 	0x202C,
 	0x202D,
 	0x202E,
+	0x2028,
+	0x2029,
 	0x2066,
 	0x2067,
 	0x2068,
@@ -49,8 +52,20 @@ var docDangerousSet = func() map[rune]bool {
 	return m
 }()
 
-func stripDocDangerousUnicode(s string) string {
+// stripDocInputUnsafe removes characters that the server-side RejectControlChars
+// validator (mirrored by apiclient.rejectDangerousChars) would reject:
+//
+//  1. C0 control characters (except tab and newline) and DEL (0x7F)
+//  2. Dangerous Unicode (zero-width, Bidi controls, line/paragraph separators, BOM)
+//
+// It is applied at the write boundary so document content passes server
+// validation instead of being rejected. Tab and newline are preserved because
+// they are legitimate in document text.
+func stripDocInputUnsafe(s string) string {
 	return strings.Map(func(r rune) rune {
+		if r != '\t' && r != '\n' && (r < 0x20 || r == 0x7F) {
+			return -1
+		}
 		if docDangerousSet[r] {
 			return -1
 		}
@@ -194,7 +209,7 @@ func prepareDocJSONMLBody(cmd *cobra.Command, raw string) (string, error) {
 	if err != nil {
 		return "", fmt.Errorf("marshal normalized jsonml: %w", err)
 	}
-	return stripDocDangerousUnicode(string(out)), nil
+	return stripDocInputUnsafe(string(out)), nil
 }
 
 func prepareDocJSONMLNode(cmd *cobra.Command, rawElement string) (string, error) {
@@ -247,7 +262,7 @@ func prepareDocJSONMLNode(cmd *cobra.Command, rawElement string) (string, error)
 	if err != nil {
 		return "", fmt.Errorf("marshal normalized jsonml: %w", err)
 	}
-	return string(out), nil
+	return stripDocInputUnsafe(string(out)), nil
 }
 
 func docEmitJSONMLFixNotes(cmd *cobra.Command, notes []string) {
diff --git a/internal/helpers/doc_strip_input_unsafe_test.go b/internal/helpers/doc_strip_input_unsafe_test.go
new file mode 100644
index 00000000..20882727
--- /dev/null
+++ b/internal/helpers/doc_strip_input_unsafe_test.go
@@ -0,0 +1,82 @@
+// Copyright 2026 Alibaba Group
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package helpers
+
+import "testing"
+
+// TestStripDocInputUnsafe verifies that stripDocInputUnsafe removes exactly the
+// characters the server-side RejectControlChars validator rejects (C0 controls
+// except tab/newline, DEL, and the dangerous-Unicode set), while leaving all
+// legitimate text untouched. Offending codepoints use explicit \u / \x escapes
+// so they are unambiguous in source.
+func TestStripDocInputUnsafe(t *testing.T) {
+	cases := []struct {
+		name string
+		in   string
+		want string
+	}{
+		{
+			name: "preserves plain text",
+			in:   "正常的文档内容 with ASCII",
+			want: "正常的文档内容 with ASCII",
+		},
+		{
+			name: "keeps tab and newline",
+			in:   "标题\n\t正文",
+			want: "标题\n\t正文",
+		},
+		{
+			name: "drops C0 controls (null, SOH, CR) and DEL",
+			in:   "正文\x00内容\x01段落\x0d结尾\x7f",
+			want: "正文内容段落结尾",
+		},
+		{
+			name: "drops zero-width space/non-joiner/joiner",
+			in:   "正文\u200b内容\u200c段落\u200d结尾",
+			want: "正文内容段落结尾",
+		},
+		{
+			name: "drops bidi overrides and isolates",
+			in:   "Bidi\u202a测试\u202e结束\u2066左\u2069右",
+			want: "Bidi测试结束左右",
+		},
+		{
+			name: "drops line and paragraph separators",
+			in:   "\u2028\u2029行段",
+			want: "行段",
+		},
+		{
+			name: "drops BOM / ZWNBSP",
+			in:   "BOM\ufeff尾",
+			want: "BOM尾",
+		},
+		{
+			name: "drops mixed control and dangerous unicode",
+			in:   "混合\x00测试\u200b结尾\x7f",
+			want: "混合测试结尾",
+		},
+		{
+			name: "empty string stays empty",
+			in:   "",
+			want: "",
+		},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			if got := stripDocInputUnsafe(tc.in); got != tc.want {
+				t.Fatalf("stripDocInputUnsafe(%q) = %q, want %q", tc.in, got, tc.want)
+			}
+		})
+	}
+}