What happens if the feedback provided is vague?

The skill applies a reasonable interpretation and then uses a 'prompt-feedback-verifier' sub-agent to identify and correct any over-fitting or under-fitting.

Will it overwrite my original prompt file immediately?

No, it uses atomic replacement. It works on a temporary copy and only replaces the original file once the verification process is successful or the user accepts the output.

How many refinement iterations will it perform?

If the initial application doesn't meet quality standards, the skill will perform up to 5 automated refinement iterations to resolve issues found by the verifier.

How does this skill prevent prompt bloat?

The skill follows strict information density rules, merging new instructions into existing sections and removing redundant phrasing to keep prompts efficient.

Does this skill require specific permissions or tools?

Yes, it requires the 'Task' tool to launch a verifier sub-agent that validates the changes before they are finalized.

Prompt Feedback Refiner

Name: Prompt Feedback Refiner
Author: doodledood

bydoodledood

•

생산성 및 워크플로

Applies user feedback to AI prompts with calibrated precision while maintaining information density and preventing regressions.

The Prompt Feedback Refiner is a specialized skill for Claude Code designed to automate the iterative process of prompt engineering. It meticulously incorporates user suggestions into existing prompts using a verification-driven workflow that prevents over-fitting and information loss. By utilizing a multi-step refinement loop, the skill ensures that every change is perfectly calibrated to the user's intent, maintaining the prompt's effectiveness while adding new capabilities or fixing observed issues.

주요 기능

01Atomic file replacement ensuring data safety throughout the process

02Iterative refinement cycles (up to 5) for complex prompt updates

03Automated verification loop to detect over-fitting and regressions

048 GitHub stars

05Precision feedback incorporation using calibrated application techniques

06Information density optimization to keep prompts concise and performant

사용 사례

01Compressing verbose prompt files while adding specific new behaviors

02Updating legacy prompts with new requirements without breaking existing logic

03Refining complex system instructions based on model performance logs

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add doodledood/claude-code-plugins apply-prompt-feedback

For use in Claude.ai and ChatGPT

Download Skill