More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    paper-radar: This Week's Papers on Your Topic, Sorted

    1

    - **Keeps only what is new:** the same paper found twice becomes one entry, and revised versions of older preprints and older articles given a new issue date are left out.

    Free

    Security scanned
    paper-radar: This Week's Papers on Your Topic, Sorted

    paper-radar: This Week's Papers on Your Topic, Sorted

    Example session with this skill installed

    What came out this week on machine-learning weather forecasting? I mostly care about data-driven global and regional forecast models and how they compare with the physics-based ones.

    • Read your context and instructions
    • Compiled the paper-radar

    I searched papers from 26 Oct to 2 Nov 2026 on arXiv and OpenAlex (which covers journals and other preprint servers); both sources responded. That turned up 185 matches. I read 48 abstracts and picked

    6 to read first:

    1. Efficient probabilistic weather forecasting at quarter-degree resolution with two-sided transport attention (Astra): a global ML ensemble model at 0.25°. In WeatherBench 2 it matches GenCast's CRPS with about 10% of GenCast's training compute.
    2. Towards operational data-driven regional weather forecasting ... with HRRRCast: an ML emulator of HRRR at 3 km that has run in near-real time for a year, compared with operational HRRR and with HREF.
    3. Comparative Verification of IFS-Control, AIFS, and WeatherNext 2 Mean ... over Inner Mongolia: the two AI products have 13–16% lower temperature error than IFS, but IFS does better on heavy 6-hour rain.

    ...three more, then

    6 worth a look (two reviews of ML weather prediction versus physics-based models, WRF-JAX on GPU, an audit of how GraphCast uses its inputs) and 12 on the edge.

    The full report is at paper-radar-2026-11-02-ml-weather-forecasting/report.md. The search matches words, so papers that describe this work in other terms may be missing, and arXiv's last day or two may be thin; a rerun in a few days would catch those.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    What it does

    A single week brings hundreds of new papers, and keyword alerts list them newest first, so the ones that matter sit between dozens that only share a word. Ask your agent what came out on your topic and it:

    • Searches four sources at once: arXiv, bioRxiv and medRxiv for preprints, and OpenAlex for journals, conference papers and most other preprint servers, for papers first posted or published in the window (the last 7 days by default, up to 62).
    • Keeps only what is new: the same paper found twice becomes one entry, and revised versions of older preprints and older articles given a new issue date are left out.
    • Ranks by fit to your topic, then reads the top 40 abstracts (and any title further down that looks on the topic) and sorts them into read first, worth a look and on the edge, with one line on why for each.
    • Never invents a paper: a script checks every pick against what was fetched, so titles, authors, venues, dates and links come from the records.
    • Writes a Markdown report with the sorted list, the next 50 papers by title, and how the list was made: the phrases, every source asked and what it answered.

    On a week with nothing squarely on your topic it says so instead of padding the list.

    What a run looks like

    In our demo a fresh agent with only this skill was asked what came out this week on machine-learning weather forecasting. In 61 seconds it searched arXiv and OpenAlex, found 185 new papers, read 48 abstracts and picked six to read first, among them a 3-km data-driven regional model verified against the operational one, an ensemble model matching GenCast's skill with about a tenth of its training compute (found by skimming titles below the reading list), and a station verification of two AI forecast products against ECMWF's physics model.

    How well it works

    Scored once on 24 past weeks it was never tuned on, in nine fields: of 64 on-topic papers that a later review of the topic cites, 48 were in its top 10 (75%; newest first: 13) and 56 in the reading list of 40 your agent reads (88%; newest first: 30). The six it never found don't use the topic's words in title or abstract. Details and intervals on the docs page.

    What's included

    • SKILL.md: turning a request into a search, running it, the tiers, the report, what to tell you
    • scripts/paper_radar.py: the search and the report, one Python file, standard library only, network calls documented at the top
    • references/categories.md: arXiv, bioRxiv and medRxiv category names
    • examples/report.md and examples/scores.json: the report and the tiers from a real run
    • references/research-folder.md: the shared research/ folder convention

    Part of the Aurora Lab research pipeline

    Free, and works entirely on its own. It pairs well with bib-check (every reference in your bibliography checked against the registries).

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Research & Analysis.

    The search matches words: a paper that describes your topic in other terms can be missed or ranked low, so broader phrases help and the agent also skims titles below the reading list. The ranking finds papers; it does not judge quality, and the tiers come from abstracts only. OpenAlex has no abstract for some publishers' papers, which are then judged by title. arXiv adds papers to its search when it announces them, so the last day or two of a window can be thin. Each run reads arXiv's 600 and OpenAlex's 200 most relevant matches. OpenAlex without a key shares a small daily budget per network. Papers only on a publisher's or conference's own site without a DOI are not covered.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 14 days ago

    • Free with an Agensi account

    Accounts
    ArxivOpenalexBiorxiv
    Software
    Python 3.10+Internet access

    Python uses only the standard library. It reaches arXiv, OpenAlex and api.biorxiv.org for bioRxiv and medRxiv.

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables

    Allowed Hosts

    export.arxiv.org
    api.openalex.org
    api.biorxiv.org

    File Scopes

    research/radar/**
    paper-radar-*/**

    Runs one Python script that sends your search phrases, the categories and the date window, nothing else, to export.arxiv.org, api.openalex.org and api.biorxiv.org (bioRxiv and medRxiv, only when you name categories). Reads two optional environment variables: OPENALEX_API_KEY (your own OpenAlex key) and PAPER_RADAR_MAILTO (your email, sent to OpenAlex). Writes one folder per run, under research/radar/ or paper-radar-<date>-<topic>/ in your working folder: the report, the reading list, the fetched records and the agent's tiers. Reads none of your other files.

    Listed14 days ago

    What's inside

    Frequently Asked Questions

    Alerts list everything that matches a keyword, newest first, from one source. paper-radar searches four sources at once, merges duplicates, ranks by fit to your topic, and has your agent read the abstracts and tell you which few to read first, with a reason for each. On 24 held-out weeks, 75% of the papers that later reviews cite were in its top 10, against 20% newest first.

    Yes. The search matches words, so a paper that describes your topic in other terms can be missed or ranked low; broader phrases help, and your agent also skims the titles below the reading list. Very new arXiv papers can take a day or two to appear in arXiv's search.

    No. The tiers say how close a paper is to your topic, judged from its abstract. The venue is shown, but a preprint can be read first.

    Any field with papers on arXiv, bioRxiv, medRxiv or in journals OpenAlex indexes. For the life sciences and medicine your agent also reads the bioRxiv and medRxiv categories that fit your topic.

    Your search phrases, the categories and the dates, to arXiv, OpenAlex and bioRxiv's API. Nothing from your files. The full list is at the top of the script.

    The search usually takes 5 to 30 seconds; each bioRxiv or medRxiv category adds a few. In our demo the whole request, reading included, took 61 seconds.

    New in Research & Analysis

    Free