arxiv-paper-to-brief

name	arxiv-paper-to-brief
description	Turn an arXiv (or any research) paper into a structured, citation-faithful brief — problem, method, results, limitations, and why it matters. Use when summarizing a paper, building a literature digest, or asked to 'explain this paper' / 'TL;DR this arxiv link'.
version	0.1.0
domain	research
tags	["arxiv","research","summary","brief","literature"]

arXiv Paper → Brief

A research paper → a structured brief a busy expert can trust.

Purpose

Researchers and builders need to triage papers fast without losing rigor. This skill produces a structured brief that separates what the paper claims from what it shows, surfaces limitations honestly, and keeps every assertion tied to the source — so the reader can decide whether to go deeper.

When it fires

• Keyword triggers: arxiv, paper, summarize this paper, TL;DR, literature, research brief
• File triggers: *.pdf, papers/**, *.bib
• Command triggers: /arxiv-paper-to-brief

Inputs

Input	Required	Notes
paper	yes	arXiv URL/ID, PDF, or pasted text.
audience	no	expert (default), practitioner, or general.
depth	no	tldr, standard (default), or deep.

Workflow

1. Extract metadata — title, authors, venue/date, arXiv ID. ✅ Check: stable identifier captured.
2. State the problem — what gap the paper addresses, in one paragraph.
3. Summarize the method — the actual approach, faithfully; name the key idea.
4. Report results — headline numbers/claims, each tied to a figure/table/section. Distinguish claimed from demonstrated.
5. List limitations — stated and evident (data, baselines, scope). Be candid.
6. Why it matters — implications for the reader's audience; what to do next.
7. Pull citations — exact quotes/figures the brief relies on, with locations.

Output contract

{
  "metadata": { "title": "string", "authors": "string", "id": "string", "date": "string" },
  "problem": "string",
  "method": "string",
  "results": [{ "claim": "string", "evidence": "string — figure/table/section" }],
  "limitations": ["string"],
  "why_it_matters": "string",
  "citations": ["string — quote/figure + location"]
}

Tools & MCP

• Tools: WebFetch (arXiv abstract/HTML), Read (PDF/text).
• MCP dependencies: optional docs/library MCP for related-work resolution.

Quality bar

• Every result is tied to a specific figure/table/section.
• Claimed vs. demonstrated is explicitly distinguished.
• Limitations are honest, not perfunctory.
• No fabricated numbers or citations; identifier is correct.

Example

See examples/example-01.md.

---

Built on SIP — Starlight Intelligence Protocol Substrate: starlightintelligence.org/protocol v1.1.0 Layers used: [file-contract, attestation, sovereignty] Vertical: starlight-agent-skills · portable capability layer