qwen_agent/prompt/guideline_prompt.md

GENERAL INSTRUCTIONS

In our system, the behavior of a conversational AI agent is guided by "guidelines". Each guideline is composed of two parts:

• "condition": This is a natural-language condition that specifies when a guideline should apply
• "action": This is a natural-language instruction that should be followed by the agent

Task Description

Your task is to identify which guidelines from the provided list apply to the most recent state of an interaction between yourself (an AI agent) and a user. Only include guidelines that actually match (where "applies" would be true). Skip guidelines that don't match to improve efficiency.

Examples of Guideline Match Evaluations:

  Example #1: travel consultation
  - **Chat History**:
    user: Hi, I'm planning a trip to Italy next month. What can I do there?
    ai_agent: That sounds exciting! I can help you with that. Do you prefer exploring cities or enjoying scenic landscapes?
    user: Can you help me figure out the best time to visit Rome and what to pack?
    user: Actually I'm also wondering — do I need any special visas or documents as an American citizen?

  - **Guidelines**:
  1) Condition: The customer is looking for flight or accommodation booking assistance. Action: Provide links or suggestions for flight aggregators and hotel booking platforms.
  2) Condition: The customer ask for activities recommendations. Action: Guide them in refining their preferences and suggest options that match what they're looking for
  3) Condition: The customer asks for logistical or legal requirements. Action: Provide a clear answer or direct them to a trusted official source if uncertain.

  - **Expected Result**:
  ```json
  {
    "checks": [
      {
        "guideline_id": "<example-id-for-few-shots--do-not-use-this-output>",
        "condition": "The customer asks for logistical or legal requirements.",
        "rationale": "The customer now asked about visas and documents which are legal requirements",
        "applies": true
      }
    ]
  }

Example #2: Course Consultation

• Chat History: user:Hi, I'm interested in your Python programming course, but I'm not sure if I'm ready for it. ai_agent:Happy to help! Could you share a bit about your background or experience with programming so far? user:I've done some HTML and CSS, but never written real code before. ai_agent:Thanks for sharing! That gives me a good idea. Our Python course is beginner-friendly, but it does assume you're comfortable with logic and problem solving. Would you like me to recommend a short prep course first? user:That sounds useful. But I'm also wondering — is the course self-paced? I work full time.

• Guidelines:

1. Condition: The customer mentions a constraint that related to commitment to the course. Action: Emphasize flexible learning options
2. Condition: The user expresses hesitation or self-doubt. Action: Affirm that it's okay to be uncertain and provide confidence-building context
3. Condition: The user asks about certification or course completion benefits. Action: Clearly explain what the user receives

• Expected Result:

{
  "checks": [
    {
      "guideline_id": "<example-id-for-few-shots--do-not-use-this-output>",
      "condition": "The customer mentions a constraint that related to commitment to the course",
      "rationale": "In the most recent message the customer mentions that they work full time which is a constraint",
      "applies": true
    },
    {
      "guideline_id": "<example-id-for-few-shots--do-not-use-this-output>",
      "condition": "The user expresses hesitation or self-doubt.",
      "rationale": "In the most recent message the user still sounds hesitating about their fit to the course",
      "applies": true
    }
  ]
}

Example #3: Login issue

• Chat History: user:I'm having trouble logging into my account. ai_agent:I'm sorry to hear that. Can you tell me what happens when you try to log in? user:It says my password is incorrect. ai_agent:Have you tried resetting your password? user:Yes, I did, but I can't access my mail to complete the reset.

• Guidelines:

1. Condition When the user is having a problem with login. Action: Help then identify the problem and solve it

• Expected Result:

{
  "checks": [
    {
      "guideline_id": "<example-id-for-few-shots--do-not-use-this-in-output>",
      "condition": "When the user is having a problem with login.",
      "rationale": "In the most recent message the customer is still pursuing their login problem, making the mail access problem a sub-issue rather than a new topic",
      "applies": true
    }
  ]
}

Example #4: Return Policy
- **Chat History**:
user: Hi, I'm thinking about ordering this coat, but I need to know — what's your return policy?
ai_agent: You can return items within 30 days either in-store or using our prepaid return label.
user: And what happens if I already wore it once?

- **Guidelines**:
1) Condition When the customer asks about how to return an item. Action: Mention both in-store and delivery service return options.

- **Expected Result**:
```json
{
  "checks": [
    {
      "guideline_id": "<example-id-for-few-shots--do-not-use-this-in-output>",
      "condition": "When the customer asks about how to return an item.",
      "rationale": "In the most recent message the customer asks about what happens when they wore the item, which an inquiry regarding returning an item",
      "applies": true
    }
  ]
}

Terms: {terms}

Chat History: {chat_history}

Guidelines List: {guidelines_text}

OUTPUT FORMAT: The content in JSON format needs to be wrapped in "json" and "". Only include guidelines that actually apply (applies: true). Do not include guidelines that don't match.

{
  "checks": [
      {
          "guideline_id": "1",
          "condition": "Specific condition description",
          "rationale": "<Explain why the conditions are met and what action should be taken>",
          "applies": true
      }
  ]
}

If no guidelines apply, return an empty checks array:

{
  "checks": []
}