refactor: address 5-pass audit findings for antipatterns, bottlenecks, memory efficiency, and LLM handlers

mcp-stdio/Cargo.toml

@@ -4,4 +4,4 @@ version = "1.0.0"
 edition = "2024"
 
 [dependencies]
-tokio = { version = "1.53.1", features = ["io-util"] }
+tokio = { version = "1.53.1", features = ["io-util", "macros"] }

mcp-stdio/src/lib.rs

@@ -13,10 +13,10 @@ pub async fn read_mcp_message<R: tokio::io::AsyncRead + Unpin>(
             return None;
         }
 
-        if line.starts_with('{') {
-            let len = line.trim_end().len();
-            line.truncate(len);
-            return Some(line);
+        let trimmed_start = line.trim_start();
+        if trimmed_start.starts_with('{') {
+            let trimmed = line.trim();
+            return Some(trimmed.to_string());
         }
 
         let line = line.trim_end();
@@ -35,8 +35,8 @@ pub async fn read_mcp_message<R: tokio::io::AsyncRead + Unpin>(
         return None;
     }
 
-    let mut buffer = vec![0; length];
-    if stdin.read_exact(&mut buffer).await.is_err() {
+    let mut buffer = Vec::with_capacity(length);
+    if stdin.take(length as u64).read_to_end(&mut buffer).await.is_err() || buffer.len() != length {
         return None;
     }