Use consistent phrasing and structure, especially for introductory phrases and verbs.
Why this matters:
It improves LLM comprehension by giving it a predictable structure.
Helps the Planner and Agent runtime interpret and prioritize actions more reliably.
Ensures a uniform tone and clear intent across all custom actions, making the agent behavior more consistent.
Example:
“Generate a summary of the most recent activity…”
“Generate a response to the customer inquiry using…”
“Generate a close plan based on opportunity stage…”
NOT C. Specify the persona who will request the action
Helpful in prompt templates, but not essential for action instructions in Agentforce.
Agent actions are triggered based on topics and utterances, not persona-specific instructions.
NOT A. Provide examples of user messages that are expected to trigger the action
This is important when defining topics, but it’s not part of the action instruction.
Examples of expected utterances belong in the topic configuration, not the action itself.
The best practice when refining Einstein Copilot custom action instructions is to provide examples of user messages that are expected to trigger the action. This helps ensure that the actions are effectively executed and meet the desired outcomes by clearly defining the input and output settings.
A voting comment increases the vote count for the chosen answer by one.
Upvoting a comment with a selected answer will also increase the vote count towards that answer by one.
So if you see a comment that you already agree with, you can upvote it instead of posting a new comment.
Jason_R
2Â weeks, 3Â days agoSappaul
1Â month, 1Â week ago92b6348
4Â months ago