Arxiv paper to brief

Skill frankxai/starlight-agent-skills/skills/research/arxiv-paper-to-brief

Portable Starlight substrate skills for agent runtimes, SIS, ACOS, and cross-agent workflow activation.

Install

npx -y skills add frankxai/starlight-agent-skills --skill arxiv-paper-to-brief

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

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

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'.

SKILL.md

2.9 KB, as published. Nobody here has run it

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

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