fix(tools): accept param aliases on say, think, speak

LLMs send varying param names for the same concept:
- say/speak: accept text, message, or phrase
- think: accept thought, text, or prompt

This complements the tool alias map (name redirect) with param
flexibility so aliased calls like dialogue(text=...) -> say and
reflection(prompt=...) -> think succeed end-to-end.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: dennys246

src/maxim/tools/narrative.py

@@ -37,7 +37,8 @@ class SayTool(Tool):
     }
 
     def execute(self, **kwargs: Any) -> ToolResult:
-        text = kwargs.get("text", "")
+        # Accept common LLM param aliases for the text to say
+        text = kwargs.get("text") or kwargs.get("message") or kwargs.get("phrase") or ""
         if not text:
             return ToolResult(success=False, error="Nothing to say")
         return ToolResult(
@@ -68,7 +69,8 @@ class ThinkTool(Tool):
     }
 
     def execute(self, **kwargs: Any) -> ToolResult:
-        thought = kwargs.get("thought", "")
+        # Accept common LLM param aliases for the thought content
+        thought = kwargs.get("thought") or kwargs.get("text") or kwargs.get("prompt") or ""
         if not thought:
             return ToolResult(success=False, error="Empty thought")
         return ToolResult(

src/maxim/tools/response.py

@@ -136,9 +136,9 @@ def execute(
         Returns:
             ToolResult with success status.
         """
-        # Accept 'message' as alias for 'text' (LLMs often use 'message')
-        if not text and kwargs.get("message"):
-            text = kwargs["message"]
+        # Accept common LLM param aliases for the text to speak
+        if not text:
+            text = kwargs.get("message") or kwargs.get("phrase") or ""
         if not text or not text.strip():
             return ToolResult(
                 success=False,