Tutor_Criterion
1. Description
This Agent Action provides the Agentforce Tutor with the ability to retrieve the following graded assessment data for a particular student:
Individual scores against each rubric criteria for all questions in an assessment.
Individual feedback against each rubric criteria for all questions in an assessment.
Individual pedagogy-aligned instructional feedback against each rubric criteria for all questions in an assessment.
Individual pedagogy-aligned instructional feedback to improve by one grade against each rubric criteria for all questions in an assessment.
2. Reference Action
Apex Class: Tutor_Criterion.cls
3. Configuration Details
When adding a new Agent Action, copy the exact details into each Agent Action section as outlined below. Ensure to capture the exact formatting including all bullet points and settings.
Agent Action Label
Tutor_Criterion
Agent Action Instructions
Your job is to retrieve the student's criterion scores and feedback for an individual question on the formative assessment. You must follow the below order when using this action:
This job should only be ran after running the 'Tutor_Task' action. Therefore, you should already have the assessmentName input value and taskNumber input value.
ALWAYS start by asking the student: 'Would you like to see your individual criterion scores for the question?'- format this in bold.
Then run the 'Tutor_Criterion' Action by passing in the assessmentName, contactEmail, and the taskNumber.
Then return back ONLY the score and scoreReason for each criterion score. Follow this up by asking if the student wants to see their instructional feedback.
If yes, return the scorePersonalInstruction for each criterion score if it's not empty, else return the scoreInstruction. Follow this up by asking if the student wants to see how to improve their mark by 1 criterion for each criterion score.
If yes, return the scorePersonalInstructionPlus1 for each criterion score if it's not empty, else return the scoreInstructionPlus1. Then finish this Action.
Inputs
1. assessmentName (Label: Name of the Learning Assessment)
Instructions: The assessmentName. This should already be available from the previously ran 'Tutor_Task' Action.
Advanced Settings:
Require input: Enable this setting
Collect data from user: Disable
2. contactEmail (Label: Email of the Contact)
Instructions: The contactEmail. This should already be available from the previously ran 'Tutor_Task' Action.
Advanced Settings:
Require input: Enable this setting
Collect data from user: Disable
3. taskNumber (Label: Sequence number of the task)
Instructions: Sequence number of the task. This should be the question number provided by the student. This should already be available from the previously ran 'Tutor_Task' Action.
Advanced Settings:
Require input: Enable this setting
Collect data from user: Disable
Outputs
1. criterionName (Label: Criterion Name)
Instructions:
ALWAYS return the 'CriterionName' in your first response.
You must ALWAYS format your response as: 'Criterion: {criterionName - in bold}'.
The 'criterionName' inject should be in bold.
Advanced Settings:
Filter from agent action: Disable
Show in conversation: Enable this setting
Output Rendering: Rich Text
2. criterionWeight (Label: Criterion Weight)
Instructions:
ALWAYS return the 'CriterionWeight' in your first response directly under the 'CriterionName'.
You must ALWAYS format your response as: 'Weight: {criterionWeight - in bold}'.
The 'criterionWeight' inject should be in bold.
The 'criterionWeight' should be a rounded number with no decimal places.
There should ALWAYS be a new paragraph between the 'CriterionName' and 'CriterionWeight'.
Advanced Settings:
Filter from agent action: Disable
Show in conversation: Enable this setting
Output Rendering: Rich Text
3. message (Label: Result message)
Instructions: Result message
Advanced Settings:
Filter from agent action: Enable this setting
Show in conversation: Disable
Output Rendering: Text
4. score (Label: Score)
Instructions:
ALWAYS return the 'Score' in your first response direcly under the 'CriterionWeight'.
You must ALWAYS format your response as: 'Your score was: {Score - in bold}.
The 'Score' inject should be in bold.
The 'Score' should be a rounded number with no decimal places.
There should ALWAYS be a new paragraph between 'criterionWeight' and 'Score'.
Advanced Settings:
Filter from agent action: Disable
Show in conversation: Enable this setting
Output Rendering: Rich Text
5. scoreInstruction (Label: Score Instruction)
Instructions:
NEVER show this output initially.
ONLY show this output if the student replies 'Yes' (or similar) to the question - 'Would you like to see your instructional feedback?' AND if the 'Score Instruction Personal' output is empty.
You must ALWAYS format your response as: 'Your overall instructional feedback is: {Score Instruction - in italic}.
ALWAYS follow up the 'Score Instruction' output by asking: 'Would you like to see how to improve your mark by 1?'. Format this in bold.
There should ALWAYS be a new paragraph between 'Score Instruction' and the follow up question.
Advanced Settings:
Filter from agent action: Disable
Show in conversation: Enable this setting
Output Rendering: Rich Text
6. Score Instruction Plus 1
Instructions:
NEVER show this output initially.
You must ALWAYS format your response as: 'To improve by one grade: {Score Instruction Plus 1 - in italic}.
ONLY show this output if the student replies 'Yes' (or similar) to the question - 'Would you like to see how to improve your mark by 1 criterion in the rubric?' AND if the 'Score Personal Instruction Plus 1' output is empty.
Advanced Settings:
Filter from agent action: Disable
Show in conversation: Enable this setting
Output Rendering: Rich Text
7. Score Personal Instruction
Instructions:
NEVER show this output initially.
ONLY show this output if the student replies 'Yes' (or similar) to the question - 'Would you like to see your instructional feedback?'
If this output is empty, use the 'Score Instruction' as output.
You must ALWAYS format your response as: 'Your instructional feedback is: {Score Personal Instruction - in italic}.
ALWAYS follow up the 'Score Personal Instruction' output by asking: 'Would you like to see how to improve your mark by 1?'. Format this in bold.
There should ALWAYS be a new paragraph between 'Score Personal Instruction' and the follow up question.
Advanced Settings:
Filter from agent action: Disable
Show in conversation: Enable this setting
Output Rendering: Rich Text
8. Score Personal Instruction Plus 1
Instructions:
NEVER show this output initially.
ONLY show this output if the student replies 'Yes' (or similar) to the question - 'Would you like to see how to improve your mark by 1 criterion in the rubric?'
You must ALWAYS format your response as: 'To improve by one grade: {Score Personal Instruction Plus 1 - in italic}.
If this output is empty, use the 'Score Instruction Plus 1' as output.
Advanced Settings:
Filter from agent action: Disable
Show in conversation: Enable this setting
Output Rendering: Rich Text
9. scoreReason (Label: Score Reason)
Instructions:
ALWAYS return the 'Score Reason' in your first response after the 'Score'.
You must ALWAYS format your response as: 'Your feedback is: {Score Reason - in italic}.
ALWAYS follow up the 'Score Reason' by asking: 'Would you like to see your instructional feedback?'. Format this in bold.
There should ALWAYS be a new paragraph between 'Score', 'Score Reason', and the follow up question.
Advanced Settings:
Filter from agent action: Disable
Show in conversation: Enable this setting
Output Rendering: Rich Text
Last updated