agentleFS
TheCraigHewittTheCraigHewitt/cowork-starter-pack/skills/doc-summarize/SKILL.md

doc-summarize

Skill5 starsChanged 6 months agoMITView on GitHub

Summarizes a long document, PDF, transcript, or report into a clear one-page brief with the key takeaways, notable quotes, and any action items. Use when the user asks to summarize, TLDR, give me the short version of, or digest a document. Works on files in the project folder or uploaded directly.

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 11 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈11

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Ask the user: what's the purpose of the summary?

    line 12
  2. Very specific:

    Read context/preferences.md — citation style, format preferences

    line 13
  3. Distinctive:

    If the document is long (>30 pages or >20k words), tell the user roughly how long the summary will take

    line 14
  4. Distinctive:

    Read or fetch the document.

    line 33
  5. Distinctive:

    If there are things the document says that contradict other things the user has read (and you have that context), flag it

    line 37
  6. Distinctive:

    Do not summarize by compression ratio ("reduce to 10%").

    line 42
  7. Distinctive:

    Do not paraphrase every paragraph.

    line 43
  8. Distinctive:

    Do not editorialize beyond what the document supports.

    line 44
  9. Distinctive:

    Do not skip quotes if the document earns them.

    line 45
  10. Distinctive:

    For research papers / academic: Add a fifth section: "What the authors are wrong or unclear about."

    line 79
  11. Distinctive:

    For product specs / tech docs: Add a "dependencies and risks" section.

    line 80
  12. Distinctive:

    If the user asked for a summary to brief someone else: Write the whole thing in a voice-neutral tone, not in the user's voice.

    line 81

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Doc Summarize

Turn a 40-page PDF, a 90-minute meeting transcript, or a dense article into a one-page brief the user can actually use.

Before starting

  1. Ask the user: what's the purpose of the summary? "For my knowledge," "for a decision I'm making," "for a meeting tomorrow" — the answer changes what you emphasize.
  2. Read context/preferences.md — citation style, format preferences
  3. If the document is long (>30 pages or >20k words), tell the user roughly how long the summary will take

What to produce

A markdown file saved to /output/summary-{document-name}-{YYYY-MM-DD}.md with four sections:

  1. The 30-second version — 3–5 sentences. The core takeaway. If the user only reads this, they got the gist.
  2. Key points — 5–10 bullets. The things that matter. Each one stands alone.
  3. Notable quotes or passages — verbatim excerpts if the document is worth quoting (reports, transcripts, academic papers). Include page/timestamp references if available.
  4. Action items or implications — if the summary has implications for the user (decisions to make, things to follow up on), list them. Skip if not applicable.

Inputs

  • The document (required) — PDF, markdown, transcript, text file, article URL
  • The purpose (recommended — ask if not given)
  • Any prior summaries or notes on related documents in the project folder

How it works

  1. Read or fetch the document. If it's a URL, pull the content. If it's a PDF, extract the text.
  2. For long documents: make a pass to identify structure (sections, arguments, conclusions), then a pass to extract the key points
  3. For transcripts: identify speakers, decisions made, action items explicitly stated
  4. Draft the four sections. Dense, specific, no padding.
  5. If there are things the document says that contradict other things the user has read (and you have that context), flag it
  6. Save the file, tell the user one sentence on what surprised you or what they should look at first

What not to do

  • Do not summarize by compression ratio ("reduce to 10%"). Summarize by importance — some documents have three things that matter, others have thirty.
  • Do not paraphrase every paragraph. Skip the filler. If 80% of a report is methodology and 20% is findings, give them the findings.
  • Do not editorialize beyond what the document supports. If you disagree with the document, that's a separate deliverable.
  • Do not skip quotes if the document earns them. Some documents are worth reading for the exact language.

Example output structure

# Summary — Q1 Board Deck (Acme Corp) — 2026-04-18

## The 30-second version
Q1 was a miss on revenue (-8% vs. plan) but a beat on new logo count (+12%). The gap is pricing — new customers came in at lower ACV than plan. CEO is proposing a pricing rework in Q2 and a hiring pause through end of Q2. Board is being asked to approve both.

## Key points
- **Revenue:** $4.2M vs. $4.6M plan. Miss.
- **New logos:** 47 vs. 42 plan. Beat.
- **ACV dropped 14% YoY** — largely due to a new self-serve tier introduced in Q4
- **Net retention:** 108%, flat with Q4, below the 112% target
- **Cash:** 22 months runway, up from 19 in Q4 (thanks to reduced spend)
- **Proposed: pricing rework** — reviewing tier structure and the self-serve plan that's cannibalizing mid-market
- **Proposed: hiring pause** — through end of Q2, open reqs close today, exceptions only for existing backfills
- **Risks called out:** competitor entered the mid-market segment in Feb, churn is ticking up in the self-serve tier

## Notable quotes
> "The self-serve tier has become a revenue problem. It's pulling customers who would otherwise buy the $15K plan into the $3K plan." — CEO letter, p. 3

> "We're not cutting. We're pausing while we figure out why our growth is outpacing our monetization." — CEO letter, p. 5

## Action items / implications
- Board to vote on pricing rework (scope: tier restructure) and hiring pause — both require approval
- Finance to model the revenue impact of three pricing scenarios before next board (dates not specified in deck)
- No action items for you personally unless you're on the pricing committee

Customization notes

  • For transcripts (meetings, interviews). Replace "notable quotes" with "decisions made" and "open questions."
  • For research papers / academic. Add a fifth section: "What the authors are wrong or unclear about."
  • For product specs / tech docs. Add a "dependencies and risks" section.
  • If the user asked for a summary to brief someone else. Write the whole thing in a voice-neutral tone, not in the user's voice. They'll forward it.
  • If the document is genuinely low-signal. Tell the user. A short, honest "this document has three things in it, here they are" is better than padding it out to look substantive.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.