Suggested Tags and Examples

1. <instruction>

This tag encloses the specific instruction you'd like to give the model.

Example:

<instruction type="summary">Summarize the following text:</instruction>

2. <input>

This tag encloses the data or text you want to be acted upon.

Example:

<input>Climate change is an urgent problem...</input>

3. <context>

This tag could enclose additional context that can help the model understand the prompt better.

Example:

<context>This is for a science project.</context>

4. <options>

Use this tag to provide multiple choices to the model, primarily useful for multiple-choice questions.

Example:

<options>
  <option>A) Apple</option>
  <option>B) Banana</option>
</options>

5. <constraints>

This tag can include any constraints that you'd like to impose on the output.

Example:

<constraints max_words="50" format="plaintext"></constraints>

6. <metadata>

For adding any additional information or notes for post-processing.

Example:

<metadata author="John" date="2023-08-28"></metadata>

7. <feedback>

For specifying a loop where the model’s output is reviewed and perhaps sent back for refinement.

Example:

<feedback>Please revise if the summary is too long.</feedback>

8. <output>

To specify the desired format or type of output.

Example:

<output type="answers-only"></output>

9. <priority>

To indicate the priority level of the task, useful for queuing multiple tasks.

Example:

<priority>high</priority>

10. <timestamp>

To record when the query was made, useful for logging purposes.

Example:

<timestamp>2023-08-28T12:34:56</timestamp>

Sample XML Prompt Using Tags:

Here's how a full prompt might look using some of these tags:

<query>
  <timestamp>2023-08-28T12:34:56</timestamp>
  <metadata author="John" date="2023-08-28"></metadata>
  <priority>high</priority>
  <context>This is for a science project.</context>
  <instruction type="summary">Summarize the following text:</instruction>
  <input>Climate change is an urgent problem...</input>
  <constraints max_words="50" format="plaintext"></constraints>
  <feedback>Please revise if the summary is too long.</feedback>
  <output type="answers-only"></output>
</query>

By structuring your prompts with these tags, you can better manage, filter, and post-process the outputs you receive from the model.