Skip to content
Conduit

skill

bio-research-skills

Agent Skills for life-science research: track the literature, read what matters, reproduce the code.

Jinsong-Zhou/bio-research-skillscommunitylistedunreviewed
Use skill

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/Jinsong-Zhou/bio-research-skills itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 23, 2026

bio-research-skills

CI Live upstream checks

Agent Skills for the daily grind of life-science research — track the literature, read what matters, reproduce the code.

Install

Requires Node.js. Works with Claude Code, Cursor, Codex and 40+ other agents via the skills CLI.

npx skills add Jinsong-Zhou/bio-research-skills                              # all skills, this project
npx skills add Jinsong-Zhou/bio-research-skills -g                           # globally
npx skills add Jinsong-Zhou/bio-research-skills --skill literature-tracking  # just one

That is the whole setup. The bundled scripts are standard-library Python 3.9+ — nothing to install, nothing to configure. Where a skill wants more than that, it says so when the moment comes and works without it in the meantime.

Skills

SkillWhat it lifts off your plate
📡 literature-trackingThree disconnected firehoses — arXiv q-bio, bioRxiv/medRxiv, PubMed. The same paper three times; alerts too broad or too narrow.
📖 paper-deep-readingOne paper, read properly. Anything can summarise; knowing whether the conclusions actually hold is the work, and it never gets written down.
🔬 code-reproductionA GitHub link ≠ the paper's numbers: dependency drift, CUDA mismatches, missing weights, an afternoon lost.

Acknowledgements

  • openags/paper-search-mcp (MIT) — the Paper record schema mirrors its field names; its _paper_unique_key is the baseline dedup rule 1 reimplements. Still the better tool for ad-hoc keyword search across 20+ sources.
  • RainerSeventeen/paper-tracker (MIT) — dedup rule 3 follows the approach in its core/dedup.py.
  • Scholar Inbox (Krishnan et al., ACL 2025 demo) — the argument for a calibrated relevance score you can threshold rather than a fixed top-N.
  • TideDra/zotero-arxiv-daily (AGPL-3.0) — studied for its recency-weighted profile ranking. Ideas only; no code copied or adapted, since AGPL is incompatible with this repository's MIT.
  • anthropics/skills (proprietary, © Anthropic) — Word and slide rendering is delegated to its docx and pptx skills. No code copied; they are prerequisites and their terms are yours to accept.
  • K-Dense-AI/scientific-agent-skills (MIT) — documenting each API's silent-failure modes in per-source references/ files is their practice, and a good one.
  • fcakyon/phd-skills (MIT) — code-reproduction deliberately stops where its reproduce skill starts, and points at it. Ideas and routing only; no code copied.
  • lllllllama/RigorPilot-Skills (MIT) — the other handoff target, for README-first reproduction of a repository that already works. Named as a prerequisite, not vendored.
  • NVIDIA-BioNeMo/Proteina-Complexa (Apache-2.0 code, NVIDIA Open Model License weights) — probe.py follows the shape of its preflight.sh, and it is the worked example throughout references/repro-hazards.md. No code copied.

License

MIT — see LICENSE.

↑↓ navigate openesc close