Performance improvements when embedding files:

- Exit early at MCP boundary if files won't fit within given context of chosen model
- Encourage claude to re-run with better context
- Check file sizes before embedding
- Drop files from older conversations when building continuations and give priority to newer files
- List and mention excluded files to Claude on return
- Improved tests
- Improved precommit prompt
- Added a new Low severity to precommit
- Improved documentation of file embedding strategy
- Refactor

tools/thinkdeep.py

@@ -143,6 +143,14 @@ class ThinkDeepTool(BaseTool):
         if updated_files is not None:
             request.files = updated_files
 
+        # MCP boundary check - STRICT REJECTION
+        if request.files:
+            file_size_check = self.check_total_file_size(request.files)
+            if file_size_check:
+                from tools.models import ToolOutput
+
+                raise ValueError(f"MCP_SIZE_CHECK:{ToolOutput(**file_size_check).model_dump_json()}")
+
         # Build context parts
         context_parts = [f"=== CLAUDE'S CURRENT ANALYSIS ===\n{current_analysis}\n=== END ANALYSIS ==="]