agentleFS
Sign inStart free
michele-orsini-71michele-orsini-71/claude-skills/summarize-article/SKILL.md

summarize-article

Skill0 starsChanged 6 months agoMITView on GitHub

Summarize an online article into a short extended abstract. Use when the user says "summarize this article", "summarize this URL", "give me an abstract of", "TL;DR this link", or pastes an article URL and asks for a summary. Fetches the page (escalating to a headless browser for dynamic content), and when the page is behind a paywall asks the user to paste the original HTML or a markdown clip.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. brpaz/agent-skills★ 1same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository10 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 9 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Call WebFetch on the URL with a prompt such as: _"Return the full article body text: title, author/byline, and every paragraph.

    line 34
  2. Distinctive:

    If the content is sufficient, go to Produce the Abstract.

    line 45
  3. Distinctive:

    When Stage 1 is insufficient because the page looks dynamic (JavaScript-rendered, empty body, "enable JavaScript"), escalate automatically …

    line 49
  4. Very specific:

    If the article is slow to appear, mcp__playwright__browser_wait_for on article text becoming visible.

    line 53
  5. Distinctive:

    If this yields the full body, go to Produce the Abstract.

    line 60
  6. Distinctive:

    If the body is still blocked after Stage 2 — a genuine paywall or login wall — do not attempt to bypass it.

    line 64
  7. Distinctive:

    Wait for the paste. Summarize whatever the user provides.

    line 71
  8. Distinctive:

    One flowing paragraph: No headings, no bullet list — prose in the style of a journal abstract.

    line 80
  9. Distinctive:

    Treat ~20 lines as a ceiling for the richest articles, never a quota to fill.

    line 84
  10. Distinctive:

    Use plain, standard English.

    line 93
  11. Distinctive:

    If the author or title cannot be determined, omit that field rather than guessing.

    line 103

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Summarize Article

Turn an article URL into a flowing extended abstract: a single paragraph whose length follows the article's substance — as short as the material allows, and only as long as it genuinely requires, never more than about 20 lines. It is written in the register of a journal abstract and must convey the author's thesis, their conclusions, and the main lines of explanation or evidence that connect the two.

The input is a URL passed as $ARGUMENTS. If $ARGUMENTS is empty, ask the user for the article URL and stop until they provide one.

Fetch Pipeline

Obtain the readable article text by escalating through these stages. Stop at the first stage that yields the full body text.

Stage 1 — WebFetch

Call WebFetch on the URL with a prompt such as: "Return the full article body text: title, author/byline, and every paragraph. Omit navigation, comments, related-links, and advertising."

Judge the result. Treat the content as insufficient when any of these hold:

  • The text is very short relative to a normal article (a few sentences only).
  • It is mostly navigation, cookie notices, or "enable JavaScript" messages.
  • It contains paywall markers: "subscribe to continue", "sign in to read", "this article is for subscribers", a truncated body ending mid-sentence.

If the content is sufficient, go to Produce the Abstract.

Stage 2 — Headless browser (automatic)

When Stage 1 is insufficient because the page looks dynamic (JavaScript-rendered, empty body, "enable JavaScript"), escalate automatically — do not ask first:

  1. mcp__playwright__browser_navigate to the URL.
  2. If the article is slow to appear, mcp__playwright__browser_wait_for on article text becoming visible.
  3. Extract the readable text. Prefer mcp__playwright__browser_evaluate to read document.body.innerText (or the main article element's innerText); use mcp__playwright__browser_snapshot if the evaluate result is not usable.
  4. mcp__playwright__browser_close when done.

If this yields the full body, go to Produce the Abstract.

Stage 3 — Ask for a paste (paywall)

If the body is still blocked after Stage 2 — a genuine paywall or login wall — do not attempt to bypass it. Ask the user:

This article appears to be behind a paywall or login wall, so I cannot fetch the full text. Please paste either the original page HTML or a markdown/plain-text clip of the article, and I will summarize that.

Wait for the paste. Summarize whatever the user provides. If they paste raw HTML, extract the article text from it before summarizing.

Produce the Abstract

From the article text, write the extended abstract:

  • One flowing paragraph. No headings, no bullet list — prose in the style of a journal abstract.
  • Length follows the article, not a fixed count. A short, single-claim piece may need only a few lines; a dense, multi-argument one may run close to 20. Treat ~20 lines as a ceiling for the richest articles, never a quota to fill. Be dense; cut filler before cutting substance.
  • Cover three things, woven together, not labeled:
    1. The author's thesis — the central claim the article advances.
    2. The author's conclusions — what they end up asserting or recommending.
    3. The main explanations — the key arguments, evidence, or steps that carry the reader from thesis to conclusion.
  • Report the author's position faithfully. Do not inject your own opinion, agreement, or critique. If the author hedges, preserve the hedge.
  • Use plain, standard English. Expand any non-obvious acronym on first use.

Begin with one short attribution line above the paragraph, then the abstract:

**<Article title>** — <author, if known>

<the extended abstract, one paragraph; length proportional to the article, ~20 lines max>

If the author or title cannot be determined, omit that field rather than guessing.

Guidelines

  • Faithful, not creative. The abstract represents the article, not your view of it. No added framing, no editorializing.
  • Escalate silently through Stages 1–2, ask only at Stage 3. The user is prompted only when a paywall genuinely blocks automated fetching.
  • One pass. Produce the abstract directly; do not ask clarifying questions about style unless the article content is itself ambiguous.

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.

Check your setup, free