Feedback

The feedback skill enables AI agents to actively solicit user input, allowing them to refine their responses and adapt behavior based on direct human evaluation. It facilitates a collaborative loop where the agent can ask for specific critiques or general satisfaction ratings after completing a task.

When to use it

• After an agent generates code, creative writing, or analysis to verify accuracy before final submission.
• During iterative design sessions where user preferences need immediate adjustment.
• To gather qualitative data on how well the agent understands complex or ambiguous instructions.
• When troubleshooting errors by asking the user to describe what went wrong in their own words.

Key capabilities

• Initiates targeted questions to gather specific feedback on recent outputs.
• Supports general sentiment checks to gauge overall user satisfaction.
• Enables dynamic conversation flow based on received critiques.

Example prompts

• "You just wrote a Python script for me. Please ask me if it meets my requirements or if any logic needs changing."
• "After summarizing this article, ask me how well you captured the main arguments and what details I might have missed."
• "I'm not sure if your explanation was clear. Ask me to rate my understanding on a scale of 1 to 5."

Tips & gotchas

Ensure the agent is configured to pause execution after receiving feedback rather than proceeding automatically. This prevents the workflow from continuing with outdated information until the user's input is processed.