progressive-summarization
/progressive-summarization
Destyluje zapisany artykuł lub notatkę warstwami: fragmenty, pogrubienia, zaznaczenia i streszczenie, by dało się ją szybko przeczytać ponownie.
- Autor
- scratchpad-sh ↗
- Repozytorium
- skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 07.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 07.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Destyluje zapisany artykuł lub notatkę warstwami: fragmenty, pogrubienia, zaznaczenia i streszczenie, by dało się ją szybko przeczytać ponownie.
Oryginalny opis (EN)
Distills a saved article or note in layers (trimmed passages, bold, highlights, and an executive summary) so it can be reread in seconds. Use when someone asks for progressive summarization, is building a second brain, wants to distill a long saved note or article, has a backlog of saved highlights they never reread and wants them distilled in layers, or wants a note made skimmable for later.
Jak działa
- When to use it
- What to ask for
- How to do it
- The format
- Example
- Pitfalls
- Saving the note
Podgląd SKILL.md
--- name: progressive-summarization description: Distills a saved article or note in layers (trimmed passages, bold, highlights, and an executive summary) so it can be reread in seconds. Use when someone asks for progressive summarization, is building a second brain, wants to distill a long saved note or article, has a backlog of saved highlights they never reread and wants them distilled in layers, or wants a note made skimmable for later. --- # Progressive summarization Progressive summarization is a way of distilling saved material in layers, each one smaller than the last, so a note can be understood at a glance months later without rereading it all. Tiago Forte described it as part of his *Building a Second Brain* approach. The layers are: 1. **Layer 1**: the saved material itself, already trimmed to the passages worth keeping. 2. **Layer 2**: the best of those passages in **bold**. 3. **Layer 3**: the best of the bold in ==highlight==. 4. **Layer 4**: a short executive summary in the person's own words, at the top. The method works because it spreads effort over time: each layer is added only when the note is used again, so only notes that keep proving useful get the full treatment. A later reader can stop at whichever layer gives them enough. ## When to use it - Someone has saved an article, transcript, or set of highlights and wants it distilled. - They have long notes they never reread because they are too long. - They are coming back to a note for a project and want it skimmable. - They ask for progressive summarization or "layers" of highlighting. For a stand-alone summary with no layers, use `summary-writer`. To turn a source into structured notes from scratch, use `source-to-notes` or `literature-notes`. To file the result, use `para-organizer`. ## What to ask for - **The material**: the article, transcript, or note. - **Which layer to stop at.** Default: all four for a short piece, layers 1 and 2 for a long one with an offer to continue, since the method says to add layers only as a note proves useful. - **What they want it for**, if a project is mentioned. The project decides what counts as "best". ## How to do it 1. **Layer 1: trim.** Keep the passages that are surprising, useful, or well put. Drop introductions, filler, ads, and repetition. Copy what you keep word for word; this layer is still the source's text. For a long article, aim to keep a fifth or less. If they give their own highlights (from a reader app), those are layer 1; do not trim them further without asking, except that an exact duplicate is kept once and mentioned. Their own comments on a highlight go on an indented line starting "My note:", so they are never read as the source's words. 2. **Layer 2: bold.** Bold the phrases that carry the main points: about one in five of the kept sentences, and only the key part of a sentence when the rest is setup. 3. **Layer 3: highlight.** Within the bold, highlight with `==text==` only the few lines that would let someone recall the whole piece. About one in five of the bold, often three to six phrases in total. On a short set of hand-picked highlights the ratios give too little: bold about a third and highlight one or two. Not every app renders `==highlight==`; if theirs does not, use bold italics. 4. **Layer 4: executive summary.** Write three to five bullets at the top in the person's own plain words, saying what the piece means for them, not only what it says. You are drafting words for them, so mark the "what it means for me" bullet `(proposed)` for them to confirm or rewrite; only they know the use. 5. Keep each layer visible. Do not delete the lower layers when you add a higher one; the point is that a reader can drill down. 6. Keep the source reference and date saved, so the note can be traced. ## The format ```markdown # [Title of the source] [Author], [publication or link] · Saved [YYYY-MM-DD] ## Summary - [Point in the person's own words] - [Point] - [What this means for them or their project] --- ## Notes [Kept passage, with **the key phrase in bold**, and ==the most important phrase highlighted==.] [Kept passage.] [Kept passage, with **a bold phrase**.] ``` ## Example Input: a pasted blog post about writing every day, around 1,200 words, including: "The goal is not to write well every day. The goal is to keep the habit alive, because a habit that survives a bad day is worth more than a perfect session once a month." and "Set the bar so low you cannot fail: one sentence counts." They want it for their newsletter. ```markdown # [?] Title not given: a blog post on writing every day [?] author, [?] link · Saved 2026-10-27 ## Summary - Daily writing is about keeping the habit, not about quality each day. - Make the minimum tiny (one sentence) so a bad day does not break the streak. - For my newsletter: a daily sentence may beat waiting for a perfect weekly draft (proposed). --- ## Notes The goal is not to write well every day. **The goal is to keep the habit alive, because ==a habit that survives a bad day is worth more than a perfect session once a month.==** **Set the bar so low you cannot fail: ==one sentence counts.==** ``` ## Pitfalls - **Bolding too much.** If half the note is bold, nothing is. Each layer should be roughly a fifth of the one below. - **Doing all four layers on everything.** The method saves effort by distilling only notes that get used. Offer to stop early. - **Rewriting the source in layers 1 to 3.** Those layers keep the source's words. Only the summary is in the person's words. - **A summary that only restates.** The top layer should say why the piece matters to the person. - **Losing the source.** Without author, link, and date, a distilled note cannot be checked or cited later. ## Saving the note If a notes app is connected (for example the Scratchpad connector, or any notes tool you can call), offer to save the result there as a new note with a clear, searchable title, or to add it to an existing note the person names. Otherwise, give the whole note as one Markdown block that pastes cleanly into any notes app. If the material came from an existing note, offer to update that note in place so the layers build on one copy.
Wersje
W bazie nie ma innych skilli o nazwie progressive-summarization.
Migawka bazy z 13.11.2026 · wersja paczki ee702b0f7713 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
Brak podobnych skilli w bazie.