Back to Details

learned-skills-index

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Learned Skills Index

已学习技能索引

This directory contains skills that have been automatically extracted from successful execution sessions. These skills represent patterns that led to quality improvements and are available for retrieval in future tasks.
本目录包含从成功执行会话中自动提取的技能。这些技能代表了能够带来质量提升的模式，可在未来任务中检索使用。

How Skills Are Learned

技能学习流程

  1. Execution: Claude Code runs a task with 
    --loop
    enabled
  2. Feedback Collection: Each iteration's quality scores and improvements are recorded
  3. Pattern Extraction: Successful patterns are extracted from quality-improving iterations
  4. Skill Generation: Patterns are compiled into a learnable skill definition
  5. Promotion Gate: Skills must meet quality thresholds before promotion
  1. 执行：Claude Code 在启用 
    --loop
    参数的情况下运行任务
  2. 反馈收集：记录每次迭代的质量评分和改进点
  3. 模式提取：从提升质量的迭代中提取成功模式
  4. 技能生成：将模式编译为可学习的技能定义
  5. 推广门槛：技能必须达到质量阈值才能被推广

Skill Status

技能状态

StatusMeaning
PendingSkill extracted but not yet promoted (needs more validation)
PromotedSkill has been validated and can be applied to new tasks
ArchivedSkill deprecated or superseded by newer learning
状态含义
Pending（待验证）技能已提取但尚未推广（需要更多验证）
Promoted（已推广）技能已通过验证，可应用于新任务
Archived（已归档）技能已弃用或被更新的学习成果取代

Quality Thresholds for Promotion

推广的质量阈值

  • Minimum quality score: 85.0
  • Minimum successful applications: 2
  • Minimum success rate: 70%
  • 最低质量评分：85.0
  • 最低成功应用次数：2
  • 最低成功率：70%

Directory Structure

目录结构

learned/
├── SKILL.md                    # This index file
├── learned-backend-auth/       # Example learned skill
│   ├── SKILL.md               # Skill definition
│   └── metadata.json          # Machine-readable metadata
└── learned-frontend-form/      # Another learned skill
    ├── SKILL.md
    └── metadata.json
learned/
├── SKILL.md                    # 本索引文件
├── learned-backend-auth/       # 示例已学习技能
│   ├── SKILL.md               # 技能定义
│   └── metadata.json          # 机器可读元数据
└── learned-frontend-form/      # 另一项已学习技能
    ├── SKILL.md
    └── metadata.json

Integration

集成方式

Learned skills are automatically retrieved based on:
  • Task description keywords
  • File types being modified
  • Domain context
To manually query learned skills:
python
from core.skill_persistence import retrieve_skills_for_task

skills = retrieve_skills_for_task(
    task_description="your task here",
    domain="backend"  # optional
)
系统会基于以下条件自动检索已学习技能：
  • 任务描述关键词
  • 正在修改的文件类型
  • 领域上下文
如需手动查询已学习技能：
python
from core.skill_persistence import retrieve_skills_for_task

skills = retrieve_skills_for_task(
    task_description="your task here",
    domain="backend"  # 可选参数
)

Safety

安全性

All learned skills include full provenance:
  • Source session ID
  • Source repository
  • Timestamp of learning
  • Quality progression
This enables rollback and audit of any behavioral changes from learned skills.
所有已学习技能都包含完整的溯源信息：
  • 源会话ID
  • 源仓库
  • 学习时间戳
  • 质量提升过程
这使得可以回滚和审核由已学习技能带来的任何行为变更。