SciAgent-Skills
Around 200 skills for Claude Code on bioinformatics, chemistry and biostatistics, from RNA-seq analysis to molecular docking
- 384 ★ on the repository
- Last commit
Extended access
Why this level
- The skills direct the agent to run Python analysis code and install scientific packages
- Connectors call external databases over the network
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/sciagent-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skills direct the agent to run Python analysis code and install scientific packages
- Connectors call external databases over the network
This is third-party code. Review the repository files before installing.
What it does
SciAgent-Skills is a library of skills for scientific computing in biology and chemistry. Each skill sits in its own folder with a SKILL.md of 300 to 550 lines: an overview, when to use it, prerequisites, a quick start, key parameters, ready recipes and a table of common errors. During planning the agent sees only the description and loads the full text when the task fits. The largest part, 65 skills, covers genomics and bioinformatics: Scanpy for single-cell RNA-seq, BioPython, pysam, the KEGG and PubMed databases. Another 26 cover structural biology and drug discovery: RDKit, AutoDock Vina, ChEMBL, PDB. The rest is split between scientific computing, cell biology, biostatistics, proteomics, systems biology, lab automation, visualization and 21 skills on scientific writing. By kind, that is 72 toolkits, 53 database connectors, 37 guides and 37 pipelines. In Claude Code the skills load as the sciagent-skills plugin and run with commands like `/sciagent-skills:scanpy-scrna-seq`, or you simply describe the task and the agent picks a fitting one. For Codex, Cursor and Windsurf the integration-templates folder holds rule templates. The author reports 92.0% on BixBench-Verified-50 against 65.3% for Claude Code without the skills.
Who it is for. For bioinformaticians, computational chemists and biostatisticians who analyze sequencing data, screen drug candidates or write papers together with Claude Code, Codex or Cursor.
Good fit when
- You have a single-cell RNA-seq count matrix and need to go from quality control to cell type annotation
- You need differential expression on a count matrix with a volcano plot for a paper
- You want ChEMBL inhibitors filtered by an IC50 threshold, screened with Lipinski rules in RDKit and docked
- For a protein structure by UniProt ID you need the AlphaFold prediction and per-region confidence
- The task calls for a Bayesian model with patient-level random effects in PyMC
Not a fit when
- You work in R: the skills are written for Python, and the Scanpy skill itself points to another route for Seurat
- You need clinical decisions or diagnostics: these are research recipes, and a person must verify the results
- You need one small skill: installing a plugin of almost 200 skills for one is wasteful, copying the folder you need is simpler
- The task is outside the natural sciences, for example web development: the skills have nothing to offer there
Example request
Load the 10X data from the filtered_feature_bc_matrix folder, filter cells by mitochondrial fraction, cluster with Leiden and find marker genes for each cluster.Limitations
Installing through the Claude Code plugin loads every skill at once, and the README describes no selective install. For other agents the repository is cloned into `.sciagent-skills` and wired in through a rules template, so automatic skill selection by description, as in Claude Code, may not apply.
The skills themselves need third-party packages: Scanpy, for instance, requires `scanpy`, `leidenalg` and `igraph`, and large datasets call for 16 GB of memory or more. The author recommends a Pixi environment and Python 3.12 or newer for the registry validation scripts. Database connectors call external services, some with their own access terms.
The CC-BY-4.0 license covers the original content, while the licenses of the described tools stay as they were. The BixBench result was obtained by the author on 50 tasks. Availability of Claude and some of the services from Russia was not checked when this card was prepared.
How to disable. In Claude Code run `/plugin uninstall sciagent-skills`. If you wired it through the templates, delete the `.sciagent-skills` folder and the block added to AGENTS.md or to the Cursor or Windsurf rules file.
FAQ
How do I install SciAgent-Skills in Claude Code?
Run `/plugin marketplace add jaechang-hits/SciAgent-Skills`, then `/plugin install sciagent-skills`. For a one-session load there is `claude --plugin-dir /path/to/SciAgent-Skills`. Check the install with `/plugin` on the Installed tab.
Do these skills work in Codex and Cursor?
Yes, the skills are plain markdown files. Clone the repository into `.sciagent-skills` in your project, then copy a template from integration-templates: AGENTS.md for Codex, cursor-rules.md into `.cursor/rules` for Cursor and windsurf-rules.md for Windsurf.
What is inside each skill?
A SKILL.md with a description for agent discovery and sections on usage, prerequisites, quick start, parameters, recipes and a troubleshooting table. Folders called references and assets may sit beside it with deeper notes and templates.
How does SciAgent-Skills differ from Scientific Agent Skills?
They are different sets for the same field. In SciAgent-Skills the skills are sorted by kind: pipeline, toolkit, database and guide, and each is indexed in registry.yaml. The author compares the set with others by skill count and BixBench result.
Where does the 92.0% BixBench figure come from?
It is the author's own measurement on BixBench-Verified-50 for Claude Code on Opus 4.6: 92.0% with the skills and 65.3% without them. The repository carries no independent check, so treat the figure as the author's claim.
Discussion
No questions or reviews yet.
Related
Skills166 skills for scientific work: bioinformatics, cheminformatics, clinical data, geospatial analysis and 100+ databases
A Google DeepMind skill collection for science tasks: access to biology and chemistry databases, literature search and structural biology
A library of agent skills for medical research: literature search and appraisal, study design, data analysis and academic writing
NVIDIA BioNeMo Agent Toolkit: skills for life sciences
NVIDIA BioNeMo Agent Toolkit
NVIDIA skills for biology and chemistry: the agent predicts protein structures, docks molecules and designs proteins
A Claude Code plugin and MCP server: over 1000 scientific tools for biology, chemistry and medicine, data-check commands and routing skills
A skill for scientific figures to Nature, Cell and Science standards: 29 figure types, data checks and vector PDF output.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| SciAgent-Skills (this card) | no | 384 | |
| Scientific Agent Skills | no | 56,313 | 166 skills for scientific work, a different set for the same field |
| Life Sciences: Anthropic's marketplace for Claude Code | yes | 642 | Anthropic's official marketplace with biology and medicine plugins |
README in short
The README opens with a benchmark table and a link to HyperLab, a web platform built on these skills. Next come the category table, five install paths by agent, the anatomy of a skill, the repository layout and six examples of skill chains. A contributor section explains how to add a skill from a template and validate the registry with a script.
README badge
Original SKILL.md
---
name: "scanpy-scrna-seq"
description: "scRNA-seq with Scanpy: QC, normalization, HVG selection, PCA, neighborhood graph, UMAP/t-SNE, Leiden clustering, markers, cell annotation, trajectory inference. Standard scRNA-seq exploration."
license: "CC-BY-4.0"
---
# Scanpy Single-Cell RNA-seq Analysis
## Overview
Scanpy is a scalable Python toolkit for analyzing single-cell RNA-seq data built on the AnnData format. This skill covers the end-to-end standard workflow: quality control, normalization, highly variable gene selection, dimensionality reduction, clustering, marker gene identification, and cell type annotation. It produces annotated datasets and publication-quality visualizations.
## When to Use
- Analyzing single-cell RNA-seq count matrices (10X Genomics, h5ad, CSV, loom)
- Performing quality control filtering on scRNA-seq datasets (mitochondrial %, gene counts)
- Running dimensionality reduction: PCA, UMAP, t-SNE
- Identifying cell clusters via Leiden community detection
- Finding differentially expressed marker genes per cluster (Wilcoxon, t-test, logistic regression)
- Annotating cell types from known marker gene panels
- Conducting trajectory inference and pseudotime analysis (PAGA, diffusion pseudotime)
- Generating publication-quality single-cell plots (dot plots, heatmaps, stacked violins)
- Comparing gene expression across experimental conditions within cell types
- Use **omics-plotting** SKILL for generic figures (volcano/heatmap/bar/KM) from exported tables; single-cell views stay in scanpy `sc.pl.*`
- Use **Seurat** (R/Bioconductor) instead for scRNA-seq analysis in an existing R workflow or when Seurat-specific assay types are required
## Prerequisites
- **Python packages**: `scanpy>=1.10`, `leidenalg`, `igraph`, `anndata`
- **Data requirements**: Gene expression count matrix (cells x genes). Common formats: 10X Cell Ranger output (`filtered_feature_bc_matrix/`), `.h5ad`, `.h5`, `.csv`
- **Environment**: Python 3.9+; 16GB+ RAM recommended for >50k cells
```bash
pip install "scanpy[leiden]" anndata
```
## Pre-flight Interview
Settle these with the user before writing any analysis code.
```yaml
decisions:
- id: D1
param: pct_counts_mt
kind: required