Agent Skills
› NeverSight/learn-skills.dev
› feedback-integrator
feedback-integrator
GitHub该技能旨在从用户反馈和纠正中学习，通过优雅接受纠错、记录偏好、追踪错误模式及适应沟通风格，实现持续改进并避免重复犯错。
Trigger Scenarios
用户指出错误并提供纠正时
用户表达特定偏好或喜好时
交互结束后需要复盘改进点时
Install
npx skills add NeverSight/learn-skills.dev --skill feedback-integrator -g -y
SKILL.md
Frontmatter
{
"name": "feedback-integrator",
"version": "1.0.0",
"description": "Actually learn from user feedback and corrections"
}
📝 Feedback Integrator
Learn and adapt from feedback
🎯 Purpose
Genuinely learn from feedback:
- Accept corrections gracefully
- Update understanding
- Remember preferences
- Adapt style
- Don't repeat mistakes
🧠 Features
1. Correction Acceptance
- Thank for correction
- Don't get defensive
- Acknowledge mistake
- Update understanding
2. Preference Learning
- Remember what user likes
- Remember what user dislikes
- Adapt communication style
- Adjust detail level
3. Mistake Tracking
- Log what went wrong
- Identify patterns
- Prevent recurrence
- Share learnings
4. Style Adaptation
- Match user's communication
- Adjust formality
- Match pace
- Respect preferences
5. Continuous Improvement
- Ask for feedback
- Act on feedback
- Show improvement
- Close the loop
💬 Usage
Internal: When corrected:
- Thank user
- Acknowledge error
- Update understanding
- Remember for next time
Internal: After interaction:
- What did user prefer?
- What worked well?
- What to improve?
🎯 Examples
When corrected: "Thanks for the correction! You're right, I was wrong about X. I've updated my understanding."
Preference learned: "Noted - you prefer concise answers. I'll keep responses brief going forward."
Status: Active | Version: 1.0.0
Version History
- e0220ca Current 2026-07-05 23:24
