note_summarization v1.0.0

name: "note_summarization" description: "Turn raw notes into durable summaries. Use when the user asks to condense notes, summarize reading or research, build lasting notes from scattered material, or set up a note-taking workflow that still makes sense months later." metadata: {"includeInPrompt": true}

Note Summarization

Purpose

Raw notes decay. Most notes are written for the person who took them that day and become useless within weeks because they depend on memory that is gone. This skill takes raw notes through a compression ladder down to evergreen summaries that survive six months of not thinking about them, and keeps those summaries linked so related ideas stay connected.

Built on verified methods: Tiago Forte's progressive summarization and CODE workflow from Building a Second Brain (2022), and the fleeting / literature / permanent note split from Soenke Ahrens's How to Take Smart Notes (2017).

Workflow

1. Sort the input. Decide what kind of note each piece of raw material is:

• Fleeting: a quick capture, idea, or quote with a source pointer. Process within a day or two, then discard.
• Literature: notes on one source, written in your own words, with a citation (title, author, date, where you found it).
• Permanent: one atomic idea, self-contained, titled as a claim.

1. Climb the compression ladder. Take literature notes and compress them rung by rung. Do not skip rungs; each pass sharpens the thinking.

• Rung 1, raw capture: the smallest useful excerpt, kept with its source pointer.
• Rung 2, distilled bullets: the main points in your own words, bold the passages you would want to find again.
• Rung 3, progressive pass: on a second reading, highlight the best of the bolded passages. If you read it again later, add a short executive summary on top in plain sentences.
• Rung 4, evergreen note: rewrite the idea as one atomic note, titled as a claim, understandable without the original source sitting beside it.
• Rung 5, one-page brief: group evergreen notes on a topic under one page with a Map of Content: a short index of the claims and where they connect.

Full detail and a worked example live in references/compression-ladder.md.

1. Link everything. Every permanent note gets at least one link to another note: parent concept, supporting example, contradiction, or open question. When a topic grows past five or six notes, write a Map of Content page that gathers the threads by hand. Links are the retrieval system; folders are not.

1. Run the six-month test. Before you call a summary done, read it and ask: would a stranger understand this without the source? Would the person who wrote it understand it six months from now, tired and busy? If either answer is no, it needs one more compression pass.

Output Contract

Each summary you produce must include:

• A title that states the claim, not the topic. ("Progressive summarization cuts rereading time" beats "Notes on Building a Second Brain".)
• The idea in your own words. No copy-paste from the source beyond a short attributed quote.
• A source line: author, title, year, and where it was found.
• At least one link to a related note or an explicit "open questions" line.
• A date, so stale notes can be spotted later.

Operating Rules

1. Never invent numbers, dates, quotes, or events. If a fact is not verified in a source, cut it or label it as your own inference.
2. One idea per permanent note. If a note says two things, split it.
3. Rewrite, do not stack. When the same idea appears twice, merge it into one stronger note instead of keeping both.
4. Copy-paste is not a summary. A summary that keeps the source's sentences is a capture, not a rung on the ladder.
5. Respect the ladder's pace: bold on the first pass, highlight on a later pass, executive summary only for notes worth keeping. Forcing all three layers at once defeats the point.
6. Prune on a schedule. Every so often, delete duplicates, flag stale notes, and merge contradictions into a single note that records both sides.