diff --git a/src/interface.jl b/src/interface.jl
index 1832a81..9db5b02 100755
--- a/src/interface.jl
+++ b/src/interface.jl
@@ -2,7 +2,7 @@ module interface
     
 
 export agentReact, agentReflex,
-        addNewMessage, clearMessage, removeLatestMsg, conversation, writeEvaluationGuidlines,
+        addNewMessage, clearMessage, removeLatestMsg, conversation, writeEvaluationGuideline,
         grading, analyze, selfReflext
 
 using JSON3, DataStructures, Dates, UUIDs, HTTP
@@ -604,9 +604,12 @@ end
 
 
 function work(a::agentReflex, usermsg::String)
+  evaluationGuideline = nothing
+
   if a.thinkingmode == :new_thinking
     a.earlierConversation = conversationSummary(a)
     _ = addNewMessage(a, "user", usermsg)
+  evaluationGuideline = writeEvaluationGuidlines(a, usermsg)
   elseif a.thinkingmode == :continue_thinking
     error("continue_thinking")
     _ = addNewMessage(a, "user", usermsg)
@@ -615,6 +618,8 @@ function work(a::agentReflex, usermsg::String)
     error("undefined condition thinkingmode = $thinkingmode")
   end
 
+  
+
   while true  # Work loop
     # plan
     a.attempt += 1
@@ -644,7 +649,8 @@ function work(a::agentReflex, usermsg::String)
       return msgToUser
     end
 
-    # evaluate
+    # if 
+
 
     
 
@@ -656,6 +662,18 @@ end
 
 """
   Actor function. 
+  
+  Args:
+    a, one of ChatAgent's agent.
+    plan, a step by step plan to respond
+  
+  Return:
+    case 1) if actor complete the plan successfully.
+      actorState = "all steps done" inidicates that all step in plan were done.
+      msgToUser = nothing.
+    case 2) if actor needs to talk to user for more context
+    actorState = "chatbox"
+    msgToUser = "message from assistant to user"
 
 """
 function actor(a::agentReflex, plan::T) where {T<:AbstractString}
@@ -705,6 +723,7 @@ function actor(a::agentReflex, plan::T) where {T<:AbstractString}
       end
     else  #TODO finish all steps
       actorState = "all steps done"
+      msgToUser = nothing
       error("actor done 2")
       break
     end
@@ -719,7 +738,7 @@ end
 
 Args:
   a, one of ChatAgent's agent.
-  shorttermMemory, a short term memory that logs what happened.
+  usermsg, stimulus e.g. question, task and etc.
 
 Return:
   An evaluation guideline used to guage AI's work.
@@ -729,23 +748,12 @@ Return:
 ```jldoctest
 julia> using ChatAgent, CommUtils
 julia> agent = ChatAgent.agentReflex("Jene")
-julia> shorttermMemory = 
-        "
-        user: What's AMD latest product?
-        assistant:   Plan 1: To provide the user with information about AMD's latest product, I will search for the most recent product release from AMD.
-          1. Search for \"AMD latest product\" using wikisearch tool.
-          2. Identify the most recent product release mentioned in the search results.
-          3. Provide the user with the name of the latest product.
-          
-          Thought 1: The user wants to know about the latest AMD products, so I should use the wikisearch tool to find information on this topic.
-          Act 1: wikisearch
-          ActInput 1: \"AMD latest product\"
-          Obs 1: No info available.
+julia> usermsg = "What's AMD latest product?"
 "
-julia> evaluationGuideLine = ChatAgent.writeEvaluationRules(agent, shorttermMemory)
+julia> evaluationGuideLine = writeEvaluationGuideline(agent, usermsg)
 ```
 """
-function writeEvaluationGuidlines(a::agentReflex, shorttermMemory::T)  where {T<:AbstractString}
+function writeEvaluationGuideline(a::agentReflex, usermsg::T)  where {T<:AbstractString}
   prompt =
   """
     <|im_start|>system
@@ -754,7 +762,7 @@ function writeEvaluationGuidlines(a::agentReflex, shorttermMemory::T)  where {T<
     wikisearch: Useful for when you need to search an encyclopedia Input is keywords and not a question.
 
     Your work:
-    $shorttermMemory
+    $usermsg
     
     Your job are:
     1. Write an evaluation guideline for your work in order to be able to evaluate your respond.
@@ -910,7 +918,7 @@ julia> report =
     source or search engine to find the most recent product release. Additionally, I could ask 
     the user for more context or clarify their question to better understand what they are 
     looking for."
-julia> lesson = analyze(agent, report)
+julia> lesson = selfReflext(agent, report)
 ```
 """
 function selfReflext(a, report::T) where {T<:AbstractString}