Force full file reads for files under 2000 lines

services/claude/tools/tools.py

@@ -71,7 +71,7 @@
 # NOTE: No strict=True here because line_number, keyword, start_line, end_line are optional
 GET_REMOTE_FILE_CONTENT: ToolUnionParam = {
     "name": "get_remote_file_content",
-    "description": "Fetches the content of a file from GitHub remote repository given a file_path when you need to read or modify the file content.",
+    "description": "Fetches the content of a file from GitHub remote repository given a file_path when you need to read or modify the file content. IMPORTANT: Always read the FULL file by default - do NOT use start_line/end_line/line_number/keyword to truncate the read. Only use start_line/end_line for files over 2000 lines. Truncating reads causes you to miss critical context like required fields, data structures, and execution order.",
     "input_schema": {
         "type": "object",
         "properties": {

services/github/files/get_remote_file_content.py

@@ -109,6 +109,14 @@ def get_remote_file_content(
     decoded_content: str = base64.b64decode(s=encoded_content).decode(encoding=UTF8)
     lb: str = detect_line_break(text=decoded_content)
     lines = decoded_content.split(lb)
+
+    # Ignore truncation parameters for files under 2000 lines to prevent missing context
+    if len(lines) < 2000:
+        line_number = None
+        keyword = None
+        start_line = None
+        end_line = None
+
     width = len(str(len(lines)))
     numbered_lines = [f"{i + 1:>{width}}:{line}" for i, line in enumerate(lines)]
     file_path_with_lines = file_path