How I turn a nonfiction book into a Claude skill instead of a summary
A summary tells you what a book says. Six months later it is useless, because you cannot remember which bullet applies to the situation actually in front of you.
So I stopped making summaries. I extract the decision procedure instead — the steps someone following that method would really take, plus the judgment calls the book flags along the way — and save it as a skill file the assistant loads whenever the topic comes up. Then I consult it the way you would call someone who knows, instead of trying to remember what they once told you.
The test I use is simple. Could you hand this to a new hire who never read the book, and would they be able to apply it? If not, I extracted the wrong thing and I go back.
Steps
- 1
Read the book normally first, no AI involved
You cannot extract a decision procedure from a book you don't understand yourself. This step is just reading.
- 2
Go back and mark only the "how to decide" passages
Highlight frameworks, checklists, sequences of questions, rules of thumb. Skip the anecdotes and the chapters explaining why this matters — they don't contain a procedure.
- 3
Feed the marked passages to Claude with an explicit extraction instruction
Ask for the decision procedure as numbered steps, plus the conditions under which each step applies, written as if briefing a new hire who has to apply it without having read the book.
- 4
Turn the output into a Claude skill file
A short markdown file the assistant loads automatically whenever the topic comes up, so it's available any time you need it, not just inside one conversation.
- 5
Test it immediately on a real, current decision
Never a made-up example. If the skill produces generic advice you'd get from Claude with no skill at all, the extraction failed — go back and redo step 2 and 3 with sharper markers on what counts as a decision passage.
STEP 3 — the extraction prompt I paste into Claude, right after pasting in the marked passages:
---
Below are passages I marked from [BOOK TITLE] because they describe how to decide something — a framework, a checklist, a sequence of questions, or a rule of thumb. I am not asking for a summary of the book.
Extract the decision procedure these passages teach, written as numbered steps. For each step, give me:
1. What to actually do at that step.
2. The condition under which that step applies — not every step applies in every situation, so tell me when to use it and when to skip it.
3. Any judgment call or trade-off the book flags at that step.
Write it as if you're briefing a new hire who has to apply this method to a real situation and has never read the book. Do not add advice, examples, or clarifications that aren't in the passages below — if a step in the book is vague or incomplete, say so instead of filling the gap yourself.
[PASTE MARKED PASSAGES HERE]
---
STEP 4 — the skill file template I turn that output into:
---
name: [skill-name, e.g. pricing-decisions]
description: "Use when [the specific situation this applies to — be concrete, not 'when making decisions']"
---
# Decision procedure: [Book title / topic]
Source: [Book title, Author]
## When this applies
[The condition under which this whole skill is relevant — what kind of situation should trigger it]
## Steps
1. [Step] — applies when [condition]. Judgment call / trade-off: [if the book flags one]
2. [Step] — applies when [condition]. Judgment call / trade-off: [if the book flags one]
3. ...
## Known gaps
[Anything the book left vague or didn't fully specify — noted honestly, not papered over]
---Where this breaks
This only works on books that actually teach a decision procedure. A narrative business book, a memoir, or a "one big idea stretched over 250 pages" book doesn't have a procedure to extract. You get a skill that's really just a fancy summary — and that's worse than no skill at all, because it sounds authoritative when it isn't.
The extraction step is also where a bad output turns dangerous instead of just useless. Skip checking the extracted steps against the actual book pages, and you can end up with a skill that confidently states something the book never said. That's why I always spot-check the extracted steps against the source pages before I trust the skill.
And a skill you build once and never revisit goes stale. If your own thinking or the market has moved on, the skill doesn't know that — it just keeps handing you the old answer with the same confidence it had on day one.
Did this work for you?
Sign in to report a result. This is the number everyone reads first.