Fix UTF-8 encoding issue in screen peek

- Read screen hardcopy output as binary first
- Decode with UTF-8 using 'replace' error handling
- Fallback to latin-1 encoding if needed
- Strip ANSI escape sequences from output
- Handles mixed encodings and binary data gracefully

This fixes crashes when screen sessions contain non-UTF-8 bytes
or terminal control sequences.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

mcp_servers/screen/screen_server.py

@@ -189,9 +189,22 @@ class ScreenServer(BaseMCPServer):
             if result.returncode != 0:
                 return self.content_text(f"Failed to peek at session: {result.stderr}")
             
-            # Read the output
+            # Read the output with proper encoding handling
-            with open(tmp_path, 'r') as f:
+            try:
-                content = f.read()
+                with open(tmp_path, 'rb') as f:
+                    raw_content = f.read()
+                
+                # Try to decode with UTF-8, replacing invalid sequences
+                content = raw_content.decode('utf-8', errors='replace')
+            except Exception:
+                # Fallback to reading with latin-1 which accepts all bytes
+                with open(tmp_path, 'r', encoding='latin-1') as f:
+                    content = f.read()
+            
+            # Clean ANSI escape sequences
+            import re
+            ansi_escape = re.compile(r'\x1B(?:[@-Z\\-_]|\[[0-?]*[ -/]*[@-~])')
+            content = ansi_escape.sub('', content)
             
             # Get last N lines
             output_lines = content.strip().split('\n')