Conversation
daniellemaxwell
commented
Jul 17, 2025
| tags: ['AI Agents', 'Prompt Engineering'] | ||
| description: "We continue our exploration of AI agents with an overview of prompt engineering." | ||
| --- | ||
| AI agents have the ability to complete tasks autonomously, but without clear, well-structured prompts, they stumble. Prompt engineering helps bridge the gap between potential and performance. |
Contributor
Author
There was a problem hiding this comment.
Added intro.
daniellemaxwell
commented
Jul 17, 2025
|
|
||
| The rule of thumb? If it’s something you want the agent to always remember (e.g., “you’re a finance expert”), it typically belongs in the system prompt. If it’s request-specific (e.g., “analyze this quarter’s sales data”), that goes into the user prompt. | ||
|
|
||
| The system prompt tends to not be visible to the end user as hiding it can prevent confusion or tampering, though [showing the prompt](https://techcrunch.com/2024/08/26/anthropic-publishes-the-system-prompt-that-makes-claude-tick/) could improve transparency and trust. |
Contributor
Author
There was a problem hiding this comment.
I included a reference to Anthropic publishing their system prompts while changing the wording a bit as it feels like system prompts generally aren't made available to end users.
daniellemaxwell
commented
Jul 17, 2025
|
|
||
| The system prompt tends to not be visible to the end user as hiding it can prevent confusion or tampering, though [showing the prompt](https://techcrunch.com/2024/08/26/anthropic-publishes-the-system-prompt-that-makes-claude-tick/) could improve transparency and trust. | ||
|
|
||
| Conflicts are possible as user prompts may override system prompts in cases like persona shifts, boundary changes, or domain switches. This may cause inconsistent behavior or confusion, since models blend instructions rather than following one set over another. To address this, researchers have proposed an [instruction hierarchy](https://arxiv.org/abs/2404.13208) that teaches LLMs how to to prioritize privileged instructions. |
Contributor
Author
There was a problem hiding this comment.
Another change here by introducing the concept of instruction hierarchy without going too in-depth on the topic.
eschmier
requested changes
Jul 17, 2025
Contributor
eschmier
left a comment
eschmier
left a comment
There was a problem hiding this comment.
Nice additions @daniellemaxwell ! My only change request is to add yourself as an author :)
eschmier
approved these changes
Jul 17, 2025
Contributor
eschmier
left a comment
eschmier
left a comment
There was a problem hiding this comment.
Approved! Thanks @daniellemaxwell !
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR adds a blog post to the series on AI agents about prompt engineering. Comments have been added to point out changes made to the original post.
https://www.awesomescreenshot.com/video/42154224?key=50c595a8087ea52594680d3fadaf759d