关于 learning-documentation-skill
learning-documentation-skill 按既定分类体系将学习记录、实验结果等洞见整理到 Supabase 的 `insights` 表,供建立可搜索的归档。适用于学习、业务、家庭等文档记录。
主要功能
- 01将洞见存入 Supabase
- 02按固定分类体系整理条目
- 03以 JSON 或 Markdown 组织内容
- 04提供插入与检索指引
- 05检查重复内容及敏感数据
适用场景
- 01记录视频、文章或课程后的笔记
- 02保存实验发现以便后续搜索
- 03归档游泳成绩或招募活动
安装
复制下面的命令,在终端中运行。
$
npx skills add https:/ / github.com/ breverdbidder/ life-os --skill learning-documentation-skill安装后在 Claude Code 中输入/learning-documentation-skill确认已加载。
详细介绍
学习文档技能
系统地将学习活动、心得和知识记录到 Life OS Supabase 数据库中。
何时使用此技能
- 完成学习活动(视频、文章、课程)后
- 记录实验心得
- 存储供日后使用的参考资料
- 随时间积累知识库
- 创建可搜索的学习档案
Supabase 结构
表: insights
位置: mocerqjnksmhcjzxrewo.supabase.co
必填字段:
category:主要分类subcategory:次要分类title:简短的描述性标题content:主要内容(结构化数据使用 JSON)metadata:其他上下文信息(可选)
分类体系
主要分类
learning:
- 子分类:youtube_transcript、article、book、course、experiment
- 用途:获取一般知识
michael_swim:
- 子分类:recruiting、meet_results、training、nutrition
- 用途:Michael 的游泳相关活动
mcp_reference:
- 子分类:architecture、claude_skills、integration_patterns
- 用途:MCP 和 Claude 技术文档
business:
- 子分类:biddeed_ai、everest_capital、strategy
- 用途:商业见解和决策
adhd:
- 子分类:task_patterns、interventions、productivity
- 用途:ADHD 管理心得
family:
- 子分类:shabbat、holidays、events
- 用途:家庭活动和纪念日
YouTube 字幕文档
时机: youtube_transcript.yml 工作流完成后
格式:
{
"category": "learning",
"subcategory": "youtube_transcript",
"title": "Claude Skills Tutorial - 6 Essential Skills",
"content": {
"video_id": "thxXGxYIwUI",
"video_url": "https://youtu.be/thxXGxYIwUI",
"transcript_summary": "Tutorial on 6 Claude skills: frontend-design, domain-brainstormer, stripe-integration, content-writer, lead-research, skill-creator",
"key_takeaways": [
"Skills only load when needed (memory efficient)",
"Progressive disclosure: ~100 tokens metadata, <5k when active",
"Can build custom skills using skill-creator"
],
"action_items": [
"Deploy 6 base skills to both repos",
"Build custom foreclosure-analysis-skill",
"Create ADHD task management skill"
],
"source": "apify"
},
"metadata": {
"duration_minutes": 15,
"watched_date": "2025-12-25",
"relevance": "high"
}
}
文章/博客文档
格式:
{
"category": "learning",
"subcategory": "article",
"title": "MCP Architecture Deep Dive",
"content": {
"url": "https://example.com/mcp-architecture",
"author": "Anthropic",
"published_date": "2025-01-15",
"key_concepts": [
"Tool-based architecture",
"Progressive disclosure",
"Context injection patterns"
],
"quotes": [
"Skills are prompt templates that inject domain-specific instructions"
],
"personal_notes": "This explains why skills are more efficient than rules"
}
}
实验/测试文档
格式:
{
"category": "business",
"subcategory": "biddeed_ai",
"title": "Smart Router V5 Performance Test",
"content": {
"experiment_date": "2025-12-24",
"hypothesis": "Gemini 2.5 Flash can handle 40-55% of requests",
"methodology": "Run 100 auction analyses, track FREE vs PAID tier usage",
"results": {
"free_tier_percentage": 52,
"paid_tier_percentage": 48,
"cost_savings": "$28/month"
},
"conclusion": "Target achieved, deploying to production",
"next_steps": [
"Monitor for 1 week",
"Adjust thresholds if needed"
]
}
}
Michael 游泳文档
比赛结果:
{
"category": "michael_swim",
"subcategory": "meet_results",
"title": "Harry Meisel Meet - December 2025",
"content": {
"meet_name": "Harry Meisel Invitational",
"date": "2025-12-13",
"location": "Orlando, FL",
"results": [
{
"event": "50 Free",
"time": "21.85",
"place": 3,
"improvement": "-0.15 from previous"
},
{
"event": "100 Free",
"time": "47.92",
"place": 5,
"improvement": "-0.32 from previous"
}
],
"notes": "Strong performance, time drops across all events",
"coach_notifications": ["UF", "FSU", "NC State"]
}
}
招募活动:
{
"category": "michael_swim",
"subcategory": "recruiting",
"title": "Coach Outreach - University of Florida",
"content": {
"university": "University of Florida",
"coach_name": "Anthony Nesty",
"contact_date": "2025-12-20",
"type": "initial_email",
"response_received": false,
"follow_up_date": "2026-01-03"
}
}
Markdown 格式规范
对于 content 字段(存储 Markdown 时):
# Title
## Key Points
- Point 1
- Point 2
## Takeaways
1. First takeaway
2. Second takeaway
## Action Items
- [ ] Action 1
- [ ] Action 2
## References
- [Source 1](url)
- [Source 2](url)
保持简洁:
- 使用标题(##)划分章节
- 使用项目符号列出清单
- 使用复选框记录待办事项
- 为参考资料添加 URL
- 不要过度使用格式
插入工作流
通过 GitHub Actions(推荐)
BidDeed.AI 仓库:
workflow: .github/workflows/insert_insight.yml
inputs:
category: "mcp_reference"
subcategory: "claude_skills"
title: "Phase 1 Skills Deployed"
content: "{json_content}"
Life OS 仓库:
workflow: .github/workflows/insert_insight.yml
inputs:
category: "learning"
subcategory: "youtube_transcript"
title: "Video Title"
content: "{json_content}"
直接插入 Supabase(谨慎使用)
仅在无法使用 GitHub Actions 时:
import requests
response = requests.post(
"https://mocerqjnksmhcjzxrewo.supabase.co/rest/v1/insights",
headers={
"apikey": "YOUR_ANON_KEY",
"Content-Type": "application/json"
},
json={
"category": "learning",
"subcategory": "article",
"title": "Example Title",
"content": {"key": "value"}
}
)
检索模式
按分类搜索:
SELECT * FROM insights
WHERE category = 'learning'
ORDER BY created_at DESC
LIMIT 10
按关键词搜索:
SELECT * FROM insights
WHERE title ILIKE '%claude skills%'
OR content::text ILIKE '%claude skills%'
近期活动:
SELECT category, subcategory, title, created_at
FROM insights
WHERE created_at > NOW() - INTERVAL '7 days'
ORDER BY created_at DESC
最佳实践
应该:
- ✓ 使用一致的分类/子分类
- ✓ 编写描述性标题
- ✓ 尽可能将内容组织为 JSON
- ✓ 在适当时添加待办事项
- ✓ 添加标签,便于日后搜索
不应该:
- ✗ 重复添加条目(插入前先检查)
- ✗ 使用含糊的标题(“笔记”“杂项”)
- ✗ 存储敏感数据(API 密钥、密码)
- ✗ 覆盖已有的有价值条目
- ✗ 插入空内容或占位内容
使用示例
"Use learning-documentation-skill to log the Claude skills video transcript"
"Document experiment results for Smart Router V5"
"Log Michael's meet results from Harry Meisel"
与其他技能集成
youtube_transcript 工作流完成后:
1. Transcript generated
2. Use learning-documentation-skill
3. Log to Supabase insights table
4. Category: learning, subcategory: youtube_transcript
任务完成后:
1. Task marked COMPLETED
2. Use learning-documentation-skill
3. Log patterns/insights discovered
4. Category: adhd, subcategory: task_patterns
重要提醒
- **保持分类一致:**始终使用标准分类
- **JSON 结构:**结构化数据优于纯文本
- **待办事项:**提取并跟踪可执行的见解
- **不要重复:**插入前先搜索
- **方便未来的自己:**撰写内容时考虑日后搜索
评论
支持 Markdown