Skip to content
Get packs

Guide · Agent skill

How to use the Nature Literature Pipeline agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The nature-literature-pipeline skill gives an AI agent a repeatable routine for keeping up with new research papers: it searches several academic sources, scores candidates against a rubric, summarises the best ones, sends a digest to a messaging app and saves notes to a local folder. This guide explains what an agent skill is, how to install this one, how to ask for a daily run, and what it needs to work. By the end you will know whether it fits your workflow and how to set it up.

Nature Literature PipelineYuan1z0825/nature-skills · Apache-2.0

Quick start

  1. 1Get the skill folder. Download the ZIP from the skill page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
  2. 2Confirm your runtime. Make sure your agent runtime supports SKILL.md skills and that you can create scheduled jobs, since the skill references Hermes cron.
  3. 3Prepare delivery and archive. Set up a Feishu, Telegram or other messaging account or webhook, and choose a local folder for archival notes.
  4. 4Describe your topic. Send the agent your research area, keywords, delivery target and archive path in one message.
  5. 5Create the daily job. Ask for a daily run at a set time with about 30 candidates and the top 5 delivered.
  6. 6Test a manual run. Run the job once manually and confirm the digest arrives and a note is written to the archive.
  7. 7Review keywords monthly. Update keywords as your research direction changes, since drift lowers match quality.

What an agent skill is and what this one does

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a task matches the skill's description. This one is a literature-monitoring pipeline with five stages: search, coarse filter, fine read, delivery and archive.

The skill searches arXiv, OpenAlex, Crossref and Semantic Scholar for about 30 candidate papers. It scores each one on six weighted dimensions (topic match ×35, methodology ×20, journal quality ×15, network relevance ×10, applied value ×10, archival value ×10) and keeps the top 5. For those, it reads at abstract or full-text level and tags the source level as Full-text, Abstract only or Metadata only.

  • Search: arXiv, OpenAlex, Crossref, Semantic Scholar, with fallback to the others if Semantic Scholar is unavailable.
  • Score: six dimensions, each capped, with the total recalculated.
  • Deliver: a formatted digest to Feishu, Telegram or a similar platform.
  • Archive: DOI and arXiv ID de-duplication, then notes written to a folder.

Install the skill from the ZIP or the Claude apps

Option 1: download the ZIP from the skill's page on this directory and unzip it into your personal skills folder, ~/.claude/skills/, or into a project's .claude/skills/ folder so it applies only to that project.

Option 2: in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.

After installing, check that the folder contains SKILL.md at its top level. The package also includes references/, templates/ and README_EN.md, which the skill uses for scoring rules, note formats and the digest template.

~/.claude/skills/nature-literature-pipeline/SKILL.md

Tell the agent your research area and delivery target

Once installed, start with one message that gives the agent your topic, keywords, where to deliver digests and where to archive notes. The agent then configures keywords, the delivery target and the archive path.

Keywords can be in English and Chinese. Scoring weights and classification tiers (A-E or your own) can also be adjusted for your field. A configuration template is provided at templates/literature-push-template.md.

My research area is [X], keywords: [Y], deliver to [feishu group name], archive to [path]

Set up the daily cron job

The daily run is driven by a scheduled job. Ask the agent to create one with a time and candidate counts. The skill's cron reference covers creating, verifying and manually recovering a job. Always confirm that a test run delivers a digest and writes an archive note, because a successful create response alone is not enough.

The scheduler is Hermes cron, which runs locally. Your computer must be on at the scheduled time.

  • Verify the job exists and is enabled.
  • Run it manually once in a test mode.
  • Check delivery and the archive note separately.
Set up a daily literature push at 08:30 Beijing time, 30 candidates, top 5 delivered

Use it for ongoing monitoring, not one-off searches

Use this skill when you want a standing watch on a topic. For a single search or a citation check, the source points to the separate nature-academic-search skill, and for CNS-style citation export to nature-citation. Pipeline outputs can later be organised in Zotero.

  • Daily digest of new papers in a fixed field.
  • Shared literature feed for a lab group.
  • Weekly batch: take the last 7 days of candidates, score them and keep the top 5.
  • Building a concentrated literature review from archived notes.

What people use it for

  • A graduate student receiving a daily digest of new papers in a specific subfield.
  • A lab group keeping a shared, automatically updated literature feed in a messaging channel.
  • A researcher building a local archive of read-status notes with duplicates skipped.
  • A weekly batch that scores recent candidates and keeps a shortlist for reading.

Frequently asked questions

Is the nature-literature-pipeline skill free?

The skill is published under the Apache-2.0 licence, as listed on this page. Note that the SKILL.md front matter names an MIT licence, so check the licence file in the package before redistributing. Costs for your agent runtime and messaging platform are separate and depend on those services.

What do I need to run it?

You need an agent runtime that supports SKILL.md skills, access to the arXiv, OpenAlex, Crossref and Semantic Scholar APIs, a messaging account or webhook such as Feishu or Telegram, and a local folder for archival notes. The skill also expects Hermes cron for scheduling, and the computer must be running at the scheduled time.

Does the skill run scripts on my machine?

The listing says it runs no scripts, but it does use the network to query academic sources and to send messages. Review the package and the delivery target before enabling it.

Why does the digest sometimes say Abstract only?

The skill tags each fine-read paper with its source level: Full-text, Abstract only or Metadata only. A tag shows what the summary was based on, so check it before relying on a paper.

Can I trust the scores?

Scores come from an agent applying a rubric, and the skill itself warns that subagents may inflate scores. It validates that each dimension is capped and the total is recalculated, but you should still read the top papers yourself.

How is it different from nature-academic-search?

nature-academic-search handles one-off searches and citation checks. nature-literature-pipeline handles continuous monitoring with scoring, scheduled delivery and archiving.

What if Semantic Scholar is down?

The skill degrades gracefully by switching to OpenAlex, Crossref and arXiv, as described in its built-in safeguards.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits researchers, graduate students and lab groups who want a standing daily or weekly digest in a fixed field and are comfortable with scheduled jobs and a local notes folder. It is a poor fit for one-off searches, for setups where the computer will be off at run time, or for work that needs verified accuracy without human review.

Download the skill