165 lines
4.9 KiB
Markdown
165 lines
4.9 KiB
Markdown
---
|
|
description: Smart capture with AI-powered auto-tagging for Bases filtering
|
|
argument-hint: [content to capture]
|
|
allowed-tools:
|
|
- Skill(obsidian-vault-manager)
|
|
- SlashCommand(/gitingest)
|
|
- Bash(*)
|
|
- mcp__fetch__fetch
|
|
- mcp__MCP_DOCKER__get_file_contents
|
|
- mcp__MCP_DOCKER__list_commits
|
|
- mcp__MCP_DOCKER__get_video_info
|
|
- mcp__MCP_DOCKER__get_transcript
|
|
- mcp__obsidian-mcp-tools__create_vault_file
|
|
---
|
|
|
|
## Task
|
|
|
|
Execute the `obsidian-vault-manager` skill for universal content capture.
|
|
|
|
**Input**: `$ARGUMENTS` (YouTube URL, GitHub URL, web article, or plain text)
|
|
**Operation**: Intelligent routing with AI-powered tagging
|
|
|
|
## Process
|
|
|
|
The skill will:
|
|
1. **Analyze content type** from input
|
|
- YouTube URL → Video capture workflow
|
|
- **GitHub URL → Delegate to `/gitingest` command**
|
|
- HTTP/HTTPS URL → Article capture workflow
|
|
- Plain text → Idea capture workflow
|
|
|
|
2. **Apply AI-powered tagging** from predefined taxonomy
|
|
- Content type tags (video, idea, article, repository)
|
|
- Topic tags (2-4 relevant topics: AI, productivity, development, etc.)
|
|
- Status tags (inbox for new captures)
|
|
- Metadata tags (actionable, technical, tutorial, etc.)
|
|
|
|
3. **Create properly formatted note** using bundled templates
|
|
- Smart filename generation
|
|
- Comprehensive frontmatter
|
|
- Structured content sections
|
|
- Tag analysis and Bases filtering suggestions
|
|
|
|
## GitHub Repository Handling (IMPORTANT)
|
|
|
|
**When a GitHub URL is detected:**
|
|
|
|
1. **Extract GitHub URL** from `$ARGUMENTS`
|
|
2. **Call `/gitingest` command** using `SlashCommand` tool:
|
|
```
|
|
SlashCommand("/gitingest https://github.com/owner/repo")
|
|
```
|
|
3. **The `/gitingest` command will:**
|
|
- Use MCP Docker GitHub tools (`get_file_contents`, `list_commits`)
|
|
- Analyze repository structure and contents
|
|
- Generate comprehensive markdown with proper tagging
|
|
- Auto-save to `/Users/zorro/Documents/Obsidian/Claudecode/`
|
|
4. **Return success** - `/gitingest` handles the complete workflow
|
|
|
|
**DO NOT:**
|
|
- ❌ Manually call `mcp__gitingest__*` tools (deprecated)
|
|
- ❌ Try to analyze GitHub repos yourself
|
|
- ❌ Use the obsidian-vault-manager skill for GitHub URLs
|
|
|
|
**Example Flow:**
|
|
```
|
|
User: /capture https://github.com/anthropics/claude-code
|
|
|
|
→ Detect: GitHub URL
|
|
→ Execute: SlashCommand("/gitingest https://github.com/anthropics/claude-code")
|
|
→ Result: Complete repository analysis saved to Obsidian
|
|
```
|
|
|
|
## Content Routing
|
|
|
|
The skill automatically routes based on input:
|
|
|
|
**YouTube Videos:**
|
|
- Pattern: `youtube.com/watch?v=` or `youtu.be/`
|
|
- Fetches transcript and metadata
|
|
- Template: `templates/youtube-note-template.md`
|
|
- Tags: `[video, {topics}, inbox, {metadata}]`
|
|
|
|
**GitHub Repositories:**
|
|
- Pattern: `github.com/owner/repo`
|
|
- Uses `/gitingest` command (MCP Docker GitHub tools)
|
|
- Creates comprehensive repository analysis
|
|
- Tags: `[repository, {language}, {topics}, inbox, technical]`
|
|
|
|
**Web Articles:**
|
|
- Pattern: HTTP/HTTPS URLs (not YouTube/GitHub)
|
|
- Fetches and summarizes content
|
|
- Extracts key takeaways
|
|
- Tags: `[article, {topics}, inbox, quick-read]`
|
|
|
|
**Ideas & Thoughts:**
|
|
- Pattern: Plain text without URL
|
|
- Template: `templates/idea-template.md`
|
|
- Smart filename from content
|
|
- Tags: `[idea, {topics}, inbox, {metadata}]`
|
|
|
|
## Tag Taxonomy
|
|
|
|
All tags come from the predefined taxonomy in the skill:
|
|
|
|
### Content Type (1 tag)
|
|
video, idea, article, study-guide, repository, reference, project
|
|
|
|
### Topics (2-4 tags)
|
|
AI, Claude, Gemini, product, marketing, projects, workflow, architecture,
|
|
design, UI-UX, coding, productivity, knowledge-management, development,
|
|
learning, research, writing, tools, business, automation, data-science,
|
|
web-development, personal-growth, finance
|
|
|
|
### Status (1 tag)
|
|
inbox, processing, evergreen, published, archived, needs-review
|
|
|
|
### Metadata (0-2 tags)
|
|
high-priority, quick-read, deep-dive, technical, conceptual,
|
|
actionable, tutorial, inspiration
|
|
|
|
## Expected Output
|
|
|
|
After successful capture:
|
|
- ✅ Content analyzed and type detected
|
|
- ✅ Smart tags applied (6-8 total)
|
|
- ✅ Note created with proper filename
|
|
- ✅ Template populated with content
|
|
- ✅ Tag analysis section added
|
|
- ✅ Bases filtering suggestions included
|
|
|
|
## Examples
|
|
|
|
**YouTube video:**
|
|
```
|
|
/capture https://youtube.com/watch?v=abc123
|
|
```
|
|
→ Creates video note with transcript, learning objectives, curriculum
|
|
|
|
**GitHub repo:**
|
|
```
|
|
/capture https://github.com/anthropics/claude-code
|
|
```
|
|
→ Creates repository analysis with architecture overview
|
|
|
|
**Article:**
|
|
```
|
|
/capture https://medium.com/article-about-ai
|
|
```
|
|
→ Creates article summary with key takeaways
|
|
|
|
**Quick idea:**
|
|
```
|
|
/capture Use AI to automatically categorize notes
|
|
```
|
|
→ Creates idea note with smart filename and tags
|
|
|
|
## Integration with Bases
|
|
|
|
Tags enable powerful Bases filtering:
|
|
- `type = video AND tags contains "AI"`
|
|
- `tags contains "inbox" AND tags contains "high-priority"`
|
|
- `tags contains "actionable" AND status = "processing"`
|
|
- `type = repository AND tags contains "development"`
|