Article summarizer 2mins
Agent skills for Claude Code and Cursor: reverse-engineer Spring Boot & Grails into EARS specs, close gaps, convert to DDD/CQRS, build knowledge graphs with graphify, and craft tool-specific prompts. A living collection — more skills added over time.
npx -y skills add anarefin/ai-agent-skills --skill article-summarizer-2minsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Fast 2-minute read of any article, blog post, or web page: a tight summary, the most critical key points, and 1–2 sharp actions — all strictly from the content, delivered instantly with no back-and-forth. Use this skill when the user wants a rapid digest and says things like "quick summary", "TL;DR", "2 min version", "give me the gist", "what's the takeaway", "brief breakdown", or drops a URL and wants fast signal. Also triggers when the user just shares a URL or article text with no instruction and speed or brevity is implied. Works for any domain.
SKILL.md
3.4 KB, 655 tokens by cl100k_base, as published. Nobody here has run it
Article Lens Quick — 2-Minute Read Skill
Produce the sharpest possible digest of any article. Hard ceiling: 2 minutes of reading (~480 words at average adult reading speed). Every word earns its place or gets cut.
Step 0 — Acquire the Content
If a URL is provided:
- Call
web_fetchimmediately. Do not ask the user to paste the text. - If fetch fails (paywall, bot-block): say "Couldn't fetch the page — paste the article text and I'll run the digest." Stop.
If raw text is pasted:
- Use it directly. No
web_fetch.
If both exist: use the pasted text.
Step 1 — Pre-flight (internal, silent)
Before writing, silently note:
- Content type — opinion/essay, tutorial, announcement, research summary, case study, listicle
- Author expertise — practitioner, academic, journalist, enthusiast
- Target audience — beginner, intermediate, senior practitioner
Do not output this. Do not ask anything. Proceed directly to output.
Step 2 — Output Structure
Three sections, in order, separated by ---. No preamble, no closing.
Before the first section, one metadata line:
📖 2 min read · Article title or inferred topic
Do not calculate — always show "2 min read" as the fixed label. This is the quick variant.
🧭 Summary
3 sentences maximum. No exceptions.
Sentence 1: the core thesis — what the article argues or shows. Sentence 2: why it matters or what problem it addresses. Sentence 3: the key caveat or constraint the author acknowledges (skip if none).
Pure prose. No bullets.
🔑 Key Points
3–5 bullets. One sentence each — no second sentences.
- Highest-signal points only. If a point could be inferred from another, cut it.
- Order by importance, not article order.
- No qualifications unless they change the meaning materially.
- Strictly from the article. No external context.
⚡ Actions & Suggestions
1–2 bullets. One sentence each.
- The single most actionable takeaway from the article for a senior reader.
- A second only if it's genuinely distinct and high-value. If in doubt, cut it.
- Grounded in something the article explicitly says or demonstrates.
Hard Limits
- 480 words total maximum — count ruthlessly, cut from Key Points first if over.
- No nested bullets at any level.
- No bold except for named concepts or terms.
- No headers beyond the three section headers.
- No editorializing — every claim traces back to the article.
- Thin article (< 400 words): note it in one line before Summary, reduce Key Points to 2–3.
Edge Cases
- Non-article URL (repo, docs, landing page): adapt, note content type beside the title in the metadata line.
- Multi-language content: respond in the user's message language.