| name | ContentAnalysis | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| pack-id | danielmiessler-contentanalysis-v1.0.0 | ||||||||||
| version | 1.0.0 | ||||||||||
| author | danielmiessler | ||||||||||
| description | Content-adaptive wisdom extraction from videos, podcasts, articles, and YouTube -- dynamic sections built from what the content actually contains | ||||||||||
| type | skill | ||||||||||
| purpose-type |
|
||||||||||
| platform | claude-code | ||||||||||
| dependencies | |||||||||||
| keywords |
|
Content-adaptive wisdom extraction -- dynamic sections built from what the content actually contains, not static templates.
Traditional content extraction follows a fixed template: IDEAS, QUOTES, HABITS, FACTS, REFERENCES. Every piece of content gets the same headers regardless of what it actually contains. A programming interview gets a "HABITS" section. A geopolitical analysis gets "FACTS" that are really just opinions. The output feels mechanical and misses the real gems because the sections were decided before the content was even read.
- Static sections miss domain-specific wisdom -- a security talk has threat model insights, not generic "ideas"
- Forced categories create padding -- filling a HABITS section when the content has none
- Uniform tone reads like a committee report -- compressed info nuggets instead of genuine observations
- No depth control -- you get the same exhaustive output whether you want a quick hit or a deep dive
The fundamental issue: the extraction format should serve the content, not the other way around.
ContentAnalysis detects what wisdom domains actually exist in the content and builds custom sections around them. A programming interview gets "Programming Philosophy" and "Developer Workflow Tips." A business podcast gets "Contrarian Business Takes" and "Money Philosophy." A security talk gets "Threat Model Insights" and "Defense Strategies."
Core capabilities:
- Dynamic section detection -- Reads the content first, identifies wisdom domains, then builds sections around what is actually there
- Five depth levels -- From Instant (one killer section) to Comprehensive (10-15 sections with themes and connections)
- Conversational voice -- Bullets that sound like someone telling a friend about it, not a book report
- Quality standards -- Every bullet earns its place; no padding, no inventory lists, no committee language
- Closing sections -- One-Sentence Takeaway, If You Only Have 2 Minutes, References and Rabbit Holes (depth-dependent)
This is the next generation of extract_wisdom. The sections adapt because the content dictates them.
This pack is designed for AI-assisted installation. Give this directory to your AI and ask it to install using INSTALL.md.
What is PAI? See the PAI Project Overview.
| Component | Path | Purpose |
|---|---|---|
| Skill definition | src/SKILL.md |
Main skill routing and workflow dispatch |
| ExtractWisdom sub-skill | src/ExtractWisdom/SKILL.md |
Dynamic content extraction skill definition with full methodology |
| Extract workflow | src/ExtractWisdom/Workflows/Extract.md |
Step-by-step extraction workflow |
Summary:
- Directories: 2 (ExtractWisdom, ExtractWisdom/Workflows)
- Files: 3
- Hooks registered: 0
- Dependencies: None (works standalone, enhanced by PAI voice and writing style configuration)
This sounds similar to the original extract_wisdom which also pulls insights from content. What makes this approach different?
The original extract_wisdom uses static sections -- IDEAS, QUOTES, HABITS, FACTS, REFERENCES -- for every piece of content. ContentAnalysis reads the content first, figures out what wisdom domains are present, and builds sections around what it finds. A talk about AI agents gets "Self-Modifying Software" and "The Agent Spectrum" instead of generic IDEAS. The tone follows a specific voice standard (Level 3 conversational) that produces bullets people actually want to read, not compressed information nuggets. Depth levels let you choose between a 30-second skim and a comprehensive deep dive.
- Dynamic sections adapted to each piece of content, not static templates
- Five depth levels from Instant to Comprehensive
- Conversational tone that reads like someone telling you about it
- Every bullet passes the "Would I tweet this?" test
- Section names read like magazine headlines, not category labels
| Trigger | What Happens |
|---|---|
| "extract wisdom from [URL]" | Full-depth extraction with 5-12 dynamic sections |
| "analyze this video" | Content scan, section detection, extraction with closing sections |
| "extract wisdom (fast)" | Fast-depth extraction: 3 sections, 3 bullets each, no closing sections |
| "extract wisdom at comprehensive level" | Maximum depth: 10-15 sections plus Themes and Connections |
| "key takeaways from this podcast" | Full extraction with emphasis on actionable insights |
| "what did I miss in this article" | Full extraction focused on non-obvious insights |
The skill auto-routes all content analysis requests to the ExtractWisdom workflow.
User: "extract wisdom from [YouTube URL about AI agents]"
AI responds:
# EXTRACT WISDOM: Building AI Agents That Actually Work
> A deep dive into agent engineering with [speaker] -- from self-modifying
> software to the economics of building with AI
---
## The Death of 80% of Apps
- He wasn't trying to build self-modifying software. He just let the agent
see its own source code and it started fixing itself.
- Most apps are just a UI on a database. Agents can generate those on the fly.
The app store as we know it is already dead.
- ...
## Money Stops Mattering
- Past a certain point, money stops mattering. A cheeseburger is a cheeseburger
no matter how rich you are.
- Every VC is offering hundreds of millions. He genuinely doesn't care.
- ...
[... more dynamic sections ...]
---
## One-Sentence Takeaway
The best agents aren't programmed -- they're given access to their own source
code and a reason to improve.
## If You Only Have 2 Minutes
- Self-modifying software happened by accident, not design
- 80% of current apps will be replaced by agent-generated UIs
- ...
## References & Rabbit Holes
- **Claude Code** -- The agent framework he built everything on
- **MCP protocol** -- The tool integration layer that makes agents practical
User: "extract wisdom (fast) from this podcast episode"
AI responds:
# EXTRACT WISDOM: [Title]
> [One-line description]
## [Dynamic Section 1]
- [3 tight bullets]
## [Dynamic Section 2]
- [3 tight bullets]
## [Dynamic Section 3]
- [3 tight bullets]
No configuration required. The skill works out of the box with sensible defaults (Full depth level).
If PAI is installed, the extraction voice is calibrated from:
~/.claude/PAI/USER/WRITINGSTYLE.md-- Canonical voice definition~/.claude/PAI/USER/SKILLCUSTOMIZATIONS/ExtractWisdom/-- Skill-specific preferences
Without these files, the skill uses its built-in Level 3 conversational voice standards.
No customization needed -- the skill adapts dynamically to each piece of content.
| Customization | Location | Impact |
|---|---|---|
| Voice/tone preferences | USER/SKILLCUSTOMIZATIONS/ExtractWisdom/PREFERENCES.md |
Adjusts bullet voice and style |
| Default depth level | USER/SKILLCUSTOMIZATIONS/ExtractWisdom/PREFERENCES.md |
Changes default from Full to another level |
| Section preferences | USER/SKILLCUSTOMIZATIONS/ExtractWisdom/PREFERENCES.md |
Always-include or always-exclude section types |
Create the customization directory at:
~/.claude/PAI/USER/SKILLCUSTOMIZATIONS/ExtractWisdom/
- Original concept: Daniel Miessler -- developed as the next generation of extract_wisdom within the PAI system
- Inspired by: The limitations of static content extraction templates
- Fabric extract_wisdom -- The original static-section content extractor that inspired this dynamic approach
- PAI Writing Style -- Voice definition that calibrates extraction tone
- Agents Pack -- Multi-perspective content analysis using parallel agents
- Investigation Pack -- Research workflows that produce content needing extraction
- PAI Voice Infrastructure -- Audio narration of extraction results
- Initial release
- Dynamic section detection based on content analysis
- Five depth levels: Instant, Fast, Basic, Full, Comprehensive
- Level 3 conversational voice standard
- Closing sections: One-Sentence Takeaway, If You Only Have 2 Minutes, References and Rabbit Holes
- Comprehensive-level Themes and Connections synthesis