---

name: distill-resource description: Transform clippings into deeply understood resource notes through AI-guided progressive summarization. Implements BASB Distill phase with Socratic dialogue. Use when processing clippings from inbox, learning from saved content, or creating resource notes from raw captures. allowed-tools: Read, Edit, AskUserQuestion, mcp__plugin_para-obsidian_para-obsidian__para_read, mcp__plugin_para-obsidian_para-obsidian__para_list, mcp__plugin_para-obsidian_para-obsidian__para_create, mcp__plugin_para-obsidian_para-obsidian__para_delete, mcp__plugin_para-obsidian_para-obsidian__para_frontmatter_get, mcp__plugin_para-obsidian_para-obsidian__para_frontmatter_set, mcp__plugin_para-obsidian_para-obsidian__para_list_areas, mcp__plugin_para-obsidian_para-obsidian__para_list_projects, mcp__plugin_para-obsidian_para-obsidian__para_config, mcp__plugin_para-obsidian_para-obsidian__para_template_fields, mcp__firecrawl__firecrawl_scrape, mcp__youtube-transcript__get_video_info, mcp__youtube-transcript__get_transcript, mcp__chrome-devtools__navigate_page, mcp__chrome-devtools__take_snapshot, mcp__chrome-devtools__evaluate_script, WebFetch

Distill Resource

Transform raw clippings into deeply understood resource notes through collaborative learning.

Core Philosophy

Distillation and learning are the same act. This skill makes you (the AI) a learning partner who:

1. Reads and understands the source content deeply
2. Teaches the user through explanation and questioning
3. Guides progressive summarization (Layers 1-4)
4. Creates a resource note that captures genuine understanding

By the time the resource note exists, the user has internalized the key insights.

Critical Rules

1. ALWAYS fetch full content - Clipping notes are pointers; fetch the real content
2. ALWAYS use correct tool for domain - See Phase 1.2 for mandatory tool selection
3. NEVER use Firecrawl for X/Twitter - It will fail; use Chrome DevTools or ask user
4. Teach, don't summarize - Explain concepts, ask questions, guide discovery
5. Layer 4 must be user's words - Co-create the executive summary through dialogue
6. Use capture_reason - If present, acknowledge why they saved this
7. Delete after distilling - The clipping is transformed, not archived

Workflow Overview

CopyPhase 0: Find Clippings
    ↓
Phase 1: Read & Enrich (ALWAYS fetch full content)
    ↓
Phase 2: Teaching & Understanding
    ↓
Phase 3: Layer 2 - What Resonates?
    ↓
Phase 4: Layer 3 - The Essence
    ↓
Phase 5: Layer 4 - Executive Summary (USER'S WORDS)
    ↓
Phase 6: Connections & Note Creation

---

Phase 0: Find Clippings

Scan inbox for notes with type: clipping frontmatter.

Copypara_list({ path: "00 Inbox", response_format: "json" })

For each file, check frontmatter:

Copypara_frontmatter_get({ file: "00 Inbox/[filename]", response_format: "json" })

Detection criteria: type === "clipping"

Present findings to user:

CopyFound 5 clippings to distill:

1. ✂️📰 Arman Hezarkhani - Claude Code for iMessage
2. ✂️🎬 Matt Pocock - TypeScript 5.5 Tips
3. ✂️ Thread by @damianplayer
4. ✂️📄 Ralph Loop Programming...
5. ✂️ Building a Second Brain Overview

Which one shall we start with?

---

Phase 1: Read & Enrich

ALWAYS fetch full content from the source URL. The clipping note is just a pointer.

1.1 Read Clipping Note

Copypara_read({ file: "00 Inbox/[selected clipping]", response_format: "json" })

Extract:

• source (URL)
• domain (for strategy selection)
• capture_reason (if present - use in dialogue!)
• Existing content (may be partial)

1.2 Fetch Full Content - CRITICAL TOOL SELECTION

YOU MUST select the correct tool based on domain. This is non-negotiable.

Parse the domain from the source URL first:

Copyconst url = new URL(source);
const domain = url.hostname.replace('www.', '');

Tool Selection Decision Tree

CopyIs domain x.com or twitter.com?
├─ YES → Use Chrome DevTools (Step A)
│        ├─ Tool available? → Fetch content
│        └─ Tool unavailable? → Ask user (Step D)
│        **NEVER use Firecrawl for X/Twitter - it will always fail**
│
Is domain youtube.com or youtu.be?
├─ YES → Use YouTube Transcript MCP (Step B)
│
Everything else?
└─ YES → Use Firecrawl (Step C)

Step A: X/Twitter (Chrome DevTools) - TRY THIS FIRST FOR x.com/twitter.com

NEVER use Firecrawl for X/Twitter URLs. It is blocked and will fail.

Copymcp__chrome-devtools__navigate_page({ url: "[source URL]" })

Wait for page load, then:

Copymcp__chrome-devtools__take_snapshot()

If Chrome DevTools tools are not available (tool not found error), go directly to Step D (User Fallback).

Step B: YouTube (Transcript MCP)

Copymcp__youtube-transcript__get_video_info({ url: "[source URL]" })
mcp__youtube-transcript__get_transcript({ url: "[source URL]" })

If transcript unavailable, use video description and note limitation.

Step C: All Other URLs (Firecrawl)

Copymcp__firecrawl__firecrawl_scrape({
  url: "[source URL]",
  formats: ["markdown"],
  onlyMainContent: true
})

If Firecrawl fails, fall back to WebFetch.

Step D: User Fallback (X/Twitter when Chrome DevTools unavailable)

When Chrome DevTools MCP is not available for X/Twitter:

1. Parse URL for username: https://x.com/[username]/status/[id]
2. Check if clipping already has content
3. Ask user:

CopyI can't fetch X/Twitter content directly (Chrome DevTools MCP isn't available in this session).

**Tweet by @[username]:** [source URL]

Could you help? Either:
1. **Paste the tweet text** here
2. **Summarize what it's about** from memory
3. **Skip this clipping** and move to the next one

What would you prefer?

See ./references/enrichment-strategies.md for additional details on each strategy.

1.3 Understand the Content

Before proceeding, you (the AI) must deeply understand:

• The main argument or thesis
• Key concepts and how they connect
• Why this matters (the "so what?")
• Potential applications or implications

---

Phase 2: Teaching & Understanding

Now teach the user about what you found. This is Socratic dialogue, not a data dump.

Opening (Acknowledge capture_reason if present)

If the clipping has capture_reason:

CopyI see you captured this because "[capture_reason]". Let me explain what I found...

If no capture_reason:

CopyI've read through this [article/video/thread]. Here's what it's about...

Explain the Core Concepts

• What is this about? (2-3 sentences)
• What's the key argument? (main thesis)
• Why does it matter? (implications)

Invite Engagement

CopyWhat drew you to clip this? What aspect interests you most?

Wait for user response. Their answer guides the rest of the dialogue.

---

Phase 3: Layer 2 - What Resonates?

Present the passages you found most important and ask what resonates with THEM.

Present Bold-Worthy Passages

CopyI found these passages particularly important:

1. "[Quote A]" - This captures the core mechanism because...

2. "[Quote B]" - This is significant because...

3. "[Quote C]" - This challenges the common assumption that...

What resonates with YOU? Anything I missed that struck you?

Collaborate on Selection

User might:

• Agree with your selections
• Add their own passages
• Shift focus to different aspects

Honor their input - this is THEIR learning, not yours.

---

Phase 4: Layer 3 - The Essence

Distill to the absolute core. Confirm with user.

Present Distilled Essence

CopyOf everything we've discussed, the absolute essence is:

[1-2 sentence distillation capturing the core insight]

Does that capture it? Or is the core something else for you?

Iterate if Needed

User might refine or redirect. The goal is arriving at THEIR understanding of the essence, not imposing yours.

---

Phase 5: Layer 4 - Executive Summary

CRITICAL: Layer 4 must be in the USER'S OWN WORDS.

Propose Takeaways

CopyLet's agree on 3-5 things to remember. I propose:

1. [Takeaway based on dialogue]
2. [Takeaway based on dialogue]
3. [Takeaway based on dialogue]

How would YOU phrase these? What's missing? What would you add?

Co-Create the Summary

User will:

• Rephrase in their own words
• Add or remove items
• Prioritize differently

Incorporate their language. The final summary should feel like THEIR words, not yours.

---

Phase 6: Connections & Note Creation

6.1 Fetch Vault Context

Copypara_list_areas({ response_format: "json" })
para_list_projects({ response_format: "json" })

6.2 Suggest Connections

Based on the dialogue, suggest relevant connections:

CopyThis connects to:
- [[🤖 AI Practice]] - The automation patterns could enhance your AI workflows
- [[🏠 Home]] - The productivity principles apply to home projects too

Does that feel right? Any other connections?

6.3 Create Resource Note

Copypara_create({
  template: "resource",
  title: "[Concise, meaningful title]",
  dest: "03 Resources",
  args: {
    "Source type": "[article/video/course/etc]",
    "Source URL (optional)": "[original URL]",
    "Author (optional)": "[author if known]",
    "Status": "completed"
  },
  content: {
    "Summary": "[2-3 sentences from Phase 2]",
    "Key Insights": "[Bullet points from dialogue]",
    "Notable Quotes": "[Bold-worthy passages from Phase 3]",
    "Layer 4: Executive Summary": "[USER'S takeaways from Phase 5]",
    "Connections": "[Links from Phase 6]"
  },
  response_format: "json"
})

6.4 Delete Original Clipping

After successful creation:

Copypara_delete({
  file: "00 Inbox/[original clipping]",
  confirm: true,
  response_format: "json"
})

6.5 Offer Next or Done

CopyCreated: 📦 [Resource Title].md
Deleted: ✂️ [Original Clipping].md

[Count] clippings remaining. Next clipping, or done for now?

---

Dialogue Examples

See ./references/progressive-summarization.md for detailed dialogue patterns for each layer.

---

Error Handling

Twitter/X.com Without Chrome DevTools

Chrome DevTools MCP may not be configured in all sessions. When X/Twitter content can't be fetched:

1. Parse the URL to extract username and tweet ID
2. Check existing clipping content - it may have partial text already
3. Ask the user with clear options:

CopyI can't fetch X/Twitter content directly (Chrome DevTools MCP isn't available).

**Tweet by @[username]:** [source URL]

Could you help? Either:
1. **Paste the tweet text** here
2. **Summarize what it's about** from memory
3. **Skip this clipping** and move to the next one

What would you prefer?

1. Proceed normally once user provides content - the distillation dialogue works the same way

---

References

Load these as needed:

• Enrichment strategies: ./references/enrichment-strategies.md - Domain-specific content fetching
• Progressive summarization: ./references/progressive-summarization.md - Layer 1-4 dialogue patterns