Learn — book & study notes archive

The user is building a personal knowledge repository of what they read and study, one file per book/material, living flat under <hub>/learning/. This skill's job is capturing what they actually say — never inventing content — and keeping the archive tidy enough that it compounds instead of rotting.

The standing rule applies here without exception: when synthesizing notes, do not include anything the user didn't write. No chapter summaries pulled from a book's general reputation, no filled-in gaps. If they haven't told you something, it isn't in the file.

Workflow

1. Identify the book/material

Read <hub>/learning/index.md (create it if missing — see below) to see what's already tracked.

• If the user names the book/material explicitly, or there's exactly one book with status reading, use it — no need to ask.
• If it's genuinely ambiguous (multiple books in progress, or nothing suggests which one), ask: "Which book/material — or a new one?" and list the known titles from the index. Never guess silently.
• New title mentioned → treat as a new book (step 3 creates its file).

2. Append the note

Slug the title (the-lean-startup, atomic-habits, etc.) → <hub>/learning/<slug>.md.

Append under a ## Conclusions section, dated:

### YYYY-MM-DD
<exactly what the user wrote — their words, lightly cleaned up for readability, nothing added>

If the file doesn't have a ## Conclusions section yet, add it.

3. New book — create the file

No existing file for this title → create <hub>/learning/<slug>.md:

---
type: learning
title: <Book Title>
author: <Author, if known — omit the line if not>
status: reading
started: YYYY-MM-DD
updated: YYYY-MM-DD
tags: [learning]
---

# <Book Title>

## Conclusions

Ask for the author if the user didn't mention it and it's not obvious — don't guess a real book's author from memory either; if you're confident from common knowledge, it's still safer to state it as inferred rather than presented as something the user said (keep it out of anything framed as "their words").

4. Update the index

<hub>/learning/index.md is the at-a-glance list. Format:

---
type: learning
updated: YYYY-MM-DD
tags: [learning, index]
---

# Learning Archive

| Title | Author | Status | Last updated |
|---|---|---|---|
| [The Lean Startup](the-lean-startup.md) | Eric Ries | reading | 2026-08-20 |

On every touch (new book, new conclusion, status change) update that book's row and bump updated in both the row and the index frontmatter.

When the user says they finished a book, flip status to finished in both the book file's frontmatter and the index row.

5. Summarize mode

When the user pastes raw notes or highlights and asks for a synthesized summary (not just a one-line conclusion), write a ## Summary — YYYY-MM-DD section synthesizing only the pasted material — organize and tighten it, don't add outside context about the book. Save it to the book's file the same way as a conclusion.

What NOT to do

• Don't fabricate chapter summaries, plot points, or "what this book is about" from training knowledge — only from what the user gives you.
• Don't create subfolders under learning/ — keep it flat.
• Don't silently pick a book when more than one is ambiguous — ask.