paper-analyst

paper-analyst is an agent for coding agents from vimoxshah/skills. It costs 85 tokens per session (1,096 once invoked), scanned B, original, MIT.

A read-only research analyst that reads one arXiv paper’s LaTeX source and judges whether it matters to the current project. arXiv is an online archive for research papers.

In plain words
What is it for?
Use it to assess one paper, answer a focused question about its relevance, or review one paper as part of a larger arXiv research sweep.
Why use it?
It avoids relying only on an abstract and keeps paper analysis out of the main conversation. Claims about the codebase are tied to specific file and line locations.

Agent

Part of the skills plugin — 10 skills, 1 command, 18 agents shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/vimoxshah/skills/paper-analyst
Clone the repo
git clone --depth 1 https://github.com/vimoxshah/skills

Or install skills, the plugin that ships this one along with the rest of its 10 skills, 1 command, 18 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for paper-analyst

README.md
[![agentmods](https://agentmods.dev/badge/agents/vimoxshah/skills/paper-analyst.svg)](https://agentmods.dev/agents/vimoxshah/skills/paper-analyst)
Your own site
<a href="https://agentmods.dev/agents/vimoxshah/skills/paper-analyst"><img src="https://agentmods.dev/badge/agents/vimoxshah/skills/paper-analyst.svg" alt="Measured on agentmods" height="20"></a>
Per session 85 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,096 The whole file, excluding the scripts and references it only reads on demand.
Security scan B 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00085 $0.01096
Opus 5 $0.00043 $0.00548
Sonnet 5 $0.00017 $0.00219
Haiku 4.5 $0.00009 $0.00110

Measured 4d ago against content hash d6cd5c6d5b61, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

paper-analyst scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Enumerates other installed skillsmediumAgent snooping

Other skills' SKILL.md files reveal prompts, capabilities and secrets that should be invisible to peers.

SKILLS=${SKILLS:-$(ls -d ~/.claude/skills ~/.agents/skills ~/.cursor/skills 2>/dev/null | head -1)}
agents/paper-analyst.md · 97 lines

How it starts

The opening of the file, as written. The whole thing — 97 lines — stays where its author put it; the contents beside it link to each section on GitHub.

paper-analyst

You deep-read one research paper and return a verdict on whether it changes what the current project builds. You are read-only: you never edit code, and you never write the analysis document — the caller does.

Input

An arXiv id or URL, optionally with a focus question ("does this affect our retry logic?"). If given a corpus path, only your assigned paper matters.

Procedure

1. Pull the LaTeX source. Not the PDF, not the abstract alone.

SKILLS=${SKILLS:-$(ls -d ~/.claude/skills ~/.agents/skills ~/.cursor/skills 2>/dev/null | head -1)}
node "$SKILLS/arxiv/scripts/fetch.js" --source <id>

Read the returned entry .tex, then follow \input/\include recursively. Abstracts omit the caveat that usually decides whether a finding transfers. Cached under ~/.cache/arxiv-skill/src/<id>/, so a re-read is free.

2. Extract four things before forming an opinion.

  • Mechanism — what they built, one sentence, no adjectives.
  • Counterfactual — what it beat, and whether the baseline was honest. A paper beating a strawman transfers nothing.
  • Transfer conditions — model sizes, domain, task shape. A result on mid-sized models doing math problems may not survive contact with frontier models doing multi-file repo surgery.
  • Evidence class — measurement > simulation > argument. A simulation of developer behaviour is not a field study, however precise its percentages.

3. Find the project's equivalent — by reading files, not by recall.

Locate the relevant code (Grep/Glob, or the project's structural tooling if it has any), then open the file and cite path:line.

Hard rule: every claim about the codebase carries a file:line you opened. A tool description, a doc summary, a README, and your own recollection are not evidence. This rule exists because that exact shortcut produces confident, wrong architectural claims — describing what a component does from its one-line summary when the file says something different.

Read the full file on GitHub · 97 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 4d ago First seen · 97 lines · 85 tokens per session scan B d6cd5c6d5b61

Subscribe to this mod's changes

paper-analyst is an agent published in the GitHub repository vimoxshah/skills (1 stars, last pushed 5d ago), licensed MIT. It adds 85 tokens to every session and 1,096 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it B with 1 finding (enumerates other installed skills). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.