Prompt Guide for Cisco AI Assistant

The Cisco AI Assistant’s Prompt Guide is designed to help you interact more effectively with our AI Assistant, ensuring you get accurate, relevant, and helpful responses to your queries and commands. Your experience with Cisco AI Assistantcan be greatly enhanced by how effectively you communicate with it.

Understanding a Prompt

A prompt is a question or any text input that you provide to the Cisco AI Assistant to initiate a conversation or request information. Essentially, it's the question you pose to the AI Assistant. The way you format and construct your prompt plays a crucial role in determining the response from the AI Assistant.

Key Components of a prompt:

Clarity: Be clear and specific about what you're asking for.
Context: Provide necessary background information.
Purpose: State what you want to achieve with your prompt.

Examples of Effective Prompts


General Prompt	Effective Prompt	What’s the difference?
What are the IP addresses and ports currently being blocked?	Can you provide me with the distinct IP addresses that are currently blocked by our firewall policies?	General prompt - Without indicating the need for "both" or "all" attributes explicitly, the assistant might provide default information on either IPs or ports, not both. Effective prompt - This prompt is clear and uses the keyword "distinct" to specify the need for unique values, which aligns with the AI Assistantcapabilities.
Tell me the firewall rules, who set them, and all the changes made last month.	I need both the names and descriptions of all active firewall rules. Please include both attributes in the output.	General prompt - This is overloaded with requests and lacks clarity on whether all attributes are needed together, leading to potential confusion for the AI Assistant. Effective prompt - This clearly states the requirement for multiple attributes by using "both," ensuring the assistant understands to include all requested information.
What are the firewall rules for IP addresses X and Y, and how do I update them?	Show me a list of all firewall rules along with their corresponding actions for the past week.	General prompt - This combines questions about rules and updating procedures, which can lead to incomplete or inaccurate responses due to lack of context or specificity. Effective prompt - The is specific about the need for a list of rules and their actions, making it a straightforward request for the AI Assistant.
Give me everything but only the names.	Initial Question: What are the current firewall rules? Follow-Up Question: Can you also provide the actions associated with these rules?	General prompt - This is ambiguous and does not use the provided keywords in a manner that the AI Assistant can effectively interpret. Effective prompt - This approach helps maintain context and ensures each question is addressed accurately.
Tell me everything about the policies on my account.	I want to understand my Edge ACP access control policy, can you tell me more about it?	General prompt - This is too vague and lacks detail. The AI Assistant is unable to determine which specific policy the user is requesting information about. Effective prompt - This informs the AI Assistant the user needs details for Edge ACP access policy. The AI Assistant will respond with all the relevant details.
Show me ports, protocols, and rule counts in Edge ACP policy, biggest to smallest.	In Edge ACP policy, what ports and protocols are configured in the rules? Include the counts of the number of rules using it and sort largest to smallest.	General prompt - This lacks specificity, combining multiple complex requests without clear instructions, and assuming the AI Assistant has implicit knowledge of how to aggregate and present the data. This leads to potential misunderstandings and responses that may not meet user expectations. Effective prompt - This approach helps maintain context and provides the assistant with clear instructions.

Guidelines for Crafting Effective Prompts

By providing precise input and context, you significantly increase the chances of receiving a targeted, relevant, and useful answer from the AI Assistant

Be Specific and provide context: Draft your with relevant information, use the correct device names, policy names, etc. that could help the AI Assistantunderstand your request better.
Use Proper Syntax: While AI Assistantcan understand colloquial language, clear and grammatically correct sentences can improve response accuracy.
Clarify the Desired Output: If you have a preference for the response format (e.g., a list, a detailed explanation, tables), mention it.
Correction and Feedback: If the response doesn't meet your expectations, you can provide feedback or ask for clarification within your next .
Direct Naming Requests: Use the phrase "give me only the names" to instruct the AI Assistant to provide solely names in its response. For example, if a user wants to know the names of firewall rules or policy names without additional details, they can use the phrase 'give me only the names of firewall rules' to instruct the AI Assistant to provide solely the names in its response.
Unique Values:Employ the keyword "unique" to request unique values from the AI Assistant.
Rules and Actions: When requesting information about rules, users can specify which attributes they want to include in the response for comprehensive insights. For example, if a user wants to know about firewall rules allowing access to a specific zone, they can specify additional attributes such as the action (e.g., allow or deny) and any relevant source zones. By providing specific instructions, users can tailor the response to their exact requirements and gain deeper insights into the configuration. This approach allows users to obtain more relevant and actionable information from the AI Assistant.
Sequential Questioning: For multiple inquiries, pose them as separate, follow-up questions to enhance clarity and context, rather than combining them into a single complex .
Explicit Multi-Attribute Queries:Clearly state "Both" or "all of the following" when seeking multiple attributes; otherwise, the AI Assistant might select an attribute at random to respond to. For example, when querying about firewall rules, attributes could include details such as the rule name, description, action (e.g., allow or deny), source IP addresses, destination IP addresses, ports, protocols, etc.

In the context of multi-attribute queries, it means requesting information about multiple characteristics or properties simultaneously. For instance, a user might want to know both the names and descriptions of firewall rules, or they might be interested in the source IP addresses and destination ports of network traffic.