diff --git a/.changepacks/changepack_log_ikN3nuiD4ptMqueTJBAjI.json b/.changepacks/changepack_log_ikN3nuiD4ptMqueTJBAjI.json
new file mode 100644
index 0000000..38b69ff
--- /dev/null
+++ b/.changepacks/changepack_log_ikN3nuiD4ptMqueTJBAjI.json
@@ -0,0 +1 @@
+{"changes":{"Cargo.toml":"Patch"},"note":"Fix description issue","date":"2026-04-07T09:36:28.999142800Z"}
\ No newline at end of file
diff --git a/crates/vespera_macro/src/parser/schema/serde_attrs.rs b/crates/vespera_macro/src/parser/schema/serde_attrs.rs
index d254d9b..3750bb7 100644
--- a/crates/vespera_macro/src/parser/schema/serde_attrs.rs
+++ b/crates/vespera_macro/src/parser/schema/serde_attrs.rs
@@ -17,13 +17,14 @@ pub fn extract_doc_comment(attrs: &[syn::Attribute]) -> Option<String> {
             }) = &meta_nv.value
         {
             let line = lit_str.value();
-            // Trim leading space that rustdoc adds.
-            // Also handle `" / "` prefix that can appear when doc-comment
-            // markers leak through TokenStream → string → parse roundtrips.
+            // Strip `" / "` or `"/ "` prefixes that can appear when doc-comment
+            // markers leak through TokenStream → string → parse roundtrips,
+            // then trim any remaining whitespace.
             let trimmed = line
                 .strip_prefix(" / ")
-                .or_else(|| line.strip_prefix(' '))
-                .unwrap_or(&line);
+                .or_else(|| line.strip_prefix("/ "))
+                .unwrap_or(&line)
+                .trim();
             doc_lines.push(trimmed.to_string());
         }
     }
diff --git a/crates/vespera_macro/src/parser/schema/struct_schema.rs b/crates/vespera_macro/src/parser/schema/struct_schema.rs
index 2bee2c6..bb34eab 100644
--- a/crates/vespera_macro/src/parser/schema/struct_schema.rs
+++ b/crates/vespera_macro/src/parser/schema/struct_schema.rs
@@ -349,6 +349,31 @@ mod tests {
         }
     }
 
+    #[test]
+    fn test_parse_struct_to_schema_description_strips_slash_prefix() {
+        // When doc attributes have "/ " prefix (without leading space), descriptions should be clean.
+        // This can happen in certain TokenStream roundtrip scenarios.
+        let struct_item: syn::ItemStruct = syn::parse_str(
+            r#"
+            #[doc = "/ Struct description"]
+            struct Admin {
+                #[doc = "/ Field description"]
+                id: i32,
+            }
+        "#,
+        )
+        .unwrap();
+        let schema = parse_struct_to_schema(&struct_item, &HashSet::new(), &HashMap::new());
+        assert_eq!(
+            schema.description,
+            Some("Struct description".to_string())
+        );
+        let props = schema.properties.unwrap();
+        if let SchemaRef::Inline(id_schema) = props.get("id").unwrap() {
+            assert_eq!(id_schema.description, Some("Field description".to_string()));
+        }
+    }
+
     #[test]
     fn test_parse_struct_to_schema_with_flatten() {
         let struct_item: syn::ItemStruct = syn::parse_str(
diff --git a/crates/vespera_macro/src/route/utils.rs b/crates/vespera_macro/src/route/utils.rs
index 06bb8b4..da2c736 100644
--- a/crates/vespera_macro/src/route/utils.rs
+++ b/crates/vespera_macro/src/route/utils.rs
@@ -14,13 +14,14 @@ pub fn extract_doc_comment(attrs: &[syn::Attribute]) -> Option<String> {
             }) = &meta_nv.value
         {
             let line = lit_str.value();
-            // Trim leading space that rustdoc adds.
-            // Also handle `" / "` prefix that can appear when doc-comment
-            // markers leak through TokenStream → string → parse roundtrips.
+            // Strip `" / "` or `"/ "` prefixes that can appear when doc-comment
+            // markers leak through TokenStream → string → parse roundtrips,
+            // then trim any remaining whitespace.
             let trimmed = line
                 .strip_prefix(" / ")
-                .or_else(|| line.strip_prefix(' '))
-                .unwrap_or(&line);
+                .or_else(|| line.strip_prefix("/ "))
+                .unwrap_or(&line)
+                .trim();
             doc_lines.push(trimmed.to_string());
         }
     }