Refine Prompt

What it does

The refine-prompt skill allows AI agents to iteratively improve their own instructions based on feedback or performance metrics. It enables dynamic adjustment of prompt parameters to optimize output quality without requiring manual intervention from the user.

When to use it

• Iterative Optimization: When an initial prompt generates acceptable but not perfect results, allowing the agent to tweak variables for better accuracy.
• Context Adaptation: When task requirements change mid-session, enabling the agent to update its instructions to align with new constraints.
• Error Correction: When the agent produces hallucinations or formatting errors, triggering an automatic rewrite of the prompt to prevent recurrence.
• Complex Task Decomposition: When breaking down multi-step workflows requires adjusting the specific focus or tone of individual sub-prompts.

Key capabilities

• Analyzes current output against desired outcomes.
• Modifies prompt structure and parameters dynamically.
• Implements feedback loops for continuous improvement.
• Adjusts tone, length, and specificity based on context.

Example prompts

• "Refine your previous response to be more concise and include specific data sources."
• "Analyze why the last output missed the user's constraints and rewrite the prompt to address them."
• "Iterate on this coding task prompt to ensure better error handling in the generated code."

Tips & gotchas

Ensure you provide clear success criteria when requesting refinement, as vague goals may lead to minor adjustments rather than significant improvements. This skill is most effective when paired with a specific evaluation metric or feedback mechanism to guide the optimization process.