| name | archive-decomposer |
| description | Decompose a content archive (podcast transcripts, course material, interview corpus) into buildable Claude Code skills. The key insight — the highest-value skills are interactive decision-making frameworks that a user applies to their own situation, NOT production tools that process the archive itself. Use when you have a large transcript corpus and want to find every teachable methodology embedded in it. |
Archive Decomposer
Break a content archive into its component skills — the repeatable decision-making frameworks, step-by-step processes, and operational methodologies that are embedded in conversations but never extracted into standalone tools.
The Core Insight (Learned the Hard Way)
There are TWO categories of skills you can extract from a content archive. Most people only see the first. The second is where the real value lives.
Category 1: Production Skills (Lower Value)
Skills where Claude processes the archive itself — content repurposing, search, mining, summarization.
Examples: "Find clip-worthy moments", "Turn a segment into a Twitter thread", "Generate show notes from a transcript"
These are useful but they're commodities. Any AI can summarize. The archive is the input being processed.
Category 2: Decision-Making Skills (Higher Value)
Skills where the archive's methodology becomes an interactive coaching tool. A founder sits down with Claude and says "I'm thinking about X" — Claude walks them through a framework that a guest or host actually taught on the show, pulling from the source transcripts as evidence.
Examples:
- "I have a business idea — vet it" → Claude runs the idea through Imad's Thiel test, Hormozi's Value Equation, Sam's unit economics math, and Shaan's North Star formula
- "I want to buy a small business" → Claude walks through Sarah Moore's 7-step acquisition process with Beshore's people diligence and Wilkinson's delegation framework
- "I got an offer to buy my company" → Claude coaches the negotiation using Sam and Shaan's own exit mistakes and rules
Why Category 2 is more valuable:
- The archive becomes a KNOWLEDGE BASE for coaching, not a thing being processed
- Users can't get this from ChatGPT — it requires the specific methodology from specific episodes
- It's the reason someone would clone the repo and open it in Claude Code
- Each skill is a differentiated product, not a generic AI feature
The Test
Before scoring any skill candidate, ask: "Is this Claude processing the archive, or is this Claude coaching the user using wisdom FROM the archive?"
If processing → Category 1 (still valid, but lower priority)
If coaching → Category 2 (build these first)
What to Look For
Signal 1: Named Frameworks
A guest or host gives a methodology a NAME or presents it as a numbered system.
Search patterns:
grep -ri "framework\|mental model\|equation\|formula" transcripts/
grep -ri "step one\|step 1\|first step\|three steps\|five steps" transcripts/
grep -ri "playbook\|blueprint\|system\|method" transcripts/
grep -ri "rule of\|rules for\|the key is\|the secret is" transcripts/
What makes it a skill: The framework has defined inputs (user's situation), clear steps (a sequence to follow), and actionable output (a decision, plan, or assessment).
Example from MFM: Hormozi's Value Equation — Value = (Dream Outcome × Perceived Likelihood) / (Time Delay × Effort) — is a named framework with 4 inputs that produces a score. Directly skillable.
Signal 2: Step-by-Step Processes
Someone walks through HOW they did something in a specific, ordered sequence. Not "here's what I think about X" but "here's exactly what I did, in order."
Search patterns:
grep -ri "here's how\|here's what I do\|what I did was" transcripts/
grep -ri "first.*then.*finally\|step by step" transcripts/
grep -ri "the process\|my process\|our process" transcripts/
What makes it a skill: The steps are transferable — another founder could follow the same sequence with different inputs and get a useful result.
Signal 3: Cautionary Frameworks (What NOT to Do)
Explicit admissions of error that contain extractable rules.
Search patterns:
grep -ri "I was wrong\|my mistake was\|what I got wrong" transcripts/
grep -ri "don't.*never\|avoid\|the worst thing" transcripts/
grep -ri "I wish I had\|if I could do it over\|looking back" transcripts/
grep -ri "regret\|learned the hard way" transcripts/
What makes it a skill: The mistake + lesson becomes a checklist item or decision rule.
Signal 4: Decision Heuristics
Simple rules of thumb that compress complex decisions into actionable tests.
Search patterns:
grep -ri "rule of thumb\|heuristic\|my test is\|I always" transcripts/
grep -ri "if.*then\|whenever I\|the way I decide" transcripts/
grep -ri "no-brainer\|the trick\|shortcut" transcripts/
Signal 5: Interactive Exercises
Moments where a host or guest describes something they do WITH another person — a conversation structure, a worksheet, a diagnostic.
Search patterns:
grep -ri "I ask them\|I would ask\|the question I ask" transcripts/
grep -ri "sit down and\|walk through\|we mapped out" transcripts/
grep -ri "exercise\|worksheet\|scorecard\|audit" transcripts/
Signal 6: Guest Expertise Operationalized
A guest who is a genuine expert in a domain lays out their methodology in enough detail that it's repeatable.
The 5-Test Filter (Score Before Building)
Before scoring I×S×T, run every framework candidate through the 5-test skill-worthiness filter. This determines what FORMAT to build, not whether to build at all.
The 5 Tests (KDP Kings = Gold Standard)
| Test | What it measures | Example |
|---|
| 1. Concrete Input | Does the user bring something specific? | A keyword, a manuscript, a business idea |
| 2. Structured Deliverable | Does it produce a template output? | Pricing card, niche report, launch timeline |
| 3. Source-Specific Knowledge (REQUIRED) | Does it contain specifics an LLM wouldn't know? | BSR thresholds, Dead Zone math, ghost categories |
| 4. Ordered Workflow | Are steps ordered with dependencies? | Step 2 depends on Step 1's output |
| 5. Less-Wrong Decisions | Does it make better choices than generic LLM? | Countdown Deal loophole, category validation |
How Tests Map to Build Type
| Passes | → Build as | Example |
|---|
| All 5 | Decision Skill (gold standard) | KDP niche-scout, pricing-strategist |
| 1 + 3 (input + source knowledge) | Tutor Skill (teach → apply → teach) | Huberman dopamine-masterclass, Lenny retention-workshop |
| 3 + some of 1/2 | Lightweight Coaching Skill (questions → framework → output) | Scott Adams reframe-engine, systems-designer |
| 3 only (source knowledge) | Framework Article (reference doc) | Named frameworks, mental models, specific protocols |
| None | Skip — generic content the LLM already knows | |
Test 3 is the only hard gate. If the LLM already knows it, don't build anything. Everything else determines format.
I×S×T Scoring (After 5-Test Classification)
For candidates that pass Test 3, score to determine build priority:
| Dimension | Question | Score |
|---|
| Interactivity | Can Claude run this as a back-and-forth coaching session? | 1-5 |
| Specificity | Are there concrete steps, numbers, templates, or rules — not just general advice? | 1-5 |
| Transferability | Does this work for different users with different inputs? | 1-5 |
| Source Density | How many episodes/guests contribute methodology to this skill? | 1-5 |
Interactivity × Specificity × Transferability = Build Priority
Source Density is a bonus multiplier: if 3+ episodes contribute frameworks to the same skill, it's a composite skill — higher value, more defensible.
Build Priority 50+: Build immediately — these are the flagship skills.
Build Priority 25-49: Build when the archive is published.
Below 25: Probably a reference doc, not a skill.
Decomposition Workflow
Phase 1: Signal Survey (Automated)
Run all the grep patterns across the full transcript archive. Don't read individual transcripts yet — just collect counts and identify which files light up across multiple signal types.
Output: A heat map of transcripts ranked by signal density. The top 20-30 transcripts are your primary sources.
Phase 2: Deep Read (Targeted)
Read the top signal-dense transcripts in full. For each, extract:
### [Framework/Process Name]
- **Speaker:** Who taught this
- **Episode:** filename + title
- **Type:** Named framework / Step-by-step / Cautionary / Heuristic / Exercise / Expert method
- **The methodology:** Summarize the actual steps or rules
- **Inputs required:** What does the user need to bring?
- **Output produced:** What does the user walk away with?
- **Composable with:** Other frameworks from other episodes that address the same domain
Phase 3: Compose Skills
The best skills are composites — they stack 2-5 frameworks from different episodes into one coherent workflow. This is where the archive's value becomes greater than any single episode.
Phase 4: Write the Skill File
Each skill file should contain:
- When to Use — trigger phrases
- The framework, in order — each step with the source quote and transcript reference
- Interactive prompts — what to ask the user at each stage
- Archive search commands — grep patterns to find additional relevant episodes
- Output template — what the user gets at the end
Phase 5: Update the CLAUDE.md
Add each skill to the repo's CLAUDE.md with:
- One-line description
- Source guests
- Example trigger phrases
Category 1 Skills (Production — Build Second)
These are still valid, just lower priority. When building them, the approach is different: you need EXAMPLES of the creator's actual output, not just the methodology.
| Skill Type | What's Needed |
|---|
| Clip finder | Examples of clips that performed well + what made them clip-worthy |
| Thread writer | Actual threads from the host as style templates |
| Show notes generator | The show's actual format as a template |
| Content repurposer | Examples of how segments became social posts |
The key: Category 1 skills are built from OUTPUT EXAMPLES. Category 2 skills are built from INPUT METHODOLOGY. Different source material, different extraction approach.
Proven Results
Skills already built from MFM using this methodology:
| Skill | Composite Sources | Category |
|---|
| idea-vetter | 5 episodes (Imad, Hormozi, Sam, Shaan, Brett Adcock) | Decision-making |
| acquisition-evaluator | 4 episodes (Sarah Moore, Beshore, Wilkinson, Codie Sanchez) | Decision-making |
| exit-coach | 3 episodes (Sam, Shaan, Wilkinson) | Decision-making |
| pricing-optimizer | 3 episodes (Patrick Campbell, Sam, Shaan) | Decision-making |
| cold-outreach | 3 episodes (Sam, Sarah Moore, Shaan) | Decision-making |
| hiring-framework | 4 episodes (Hormozi, Wilkinson, Beshore, Blake) | Decision-making |
Decompose the methodology, not the content. The archive is the knowledge base, not the product. The skills are the product.