Scrna Seq Workbench — Development skill for Claude Code

scRNA-seq Workbench

An agent plugin for single-cell RNA-seq analysis, with five Skills and a shared Python runner. Use it with **Codex**, **Claude Code** or **DeepSeek Harness**. Version **0.2.1 - research preview**.

Give your assistant a count matrix, sample information and an analysis goal. The Skills guide it through inspecting your files, explaining parameter choices, running the analysis and reviewing the results with you.

Get started

1. Follow the installation tutorial: Windows/D: and Linux/macOS instructions, Python dependencies and common errors.
2. Enable the plugin in your assistant.
3. Run the 400-cell example to check your setup.
4. Follow Analyze your own data for file preparation, parameter choices and the five analysis stages.

Installing the plugin adds instructions and scripts. Its calculations use a Python environment on your computer or server. scVI and differential expression have additional dependencies described in the installation tutorial.

What it does

Skill	What you provide	What you receive
sequencing-report-review	Vendor report and sample details	A summary of sequencing metrics, source evidence and missing files
scrna-qc	UMI counts, species and sample metadata	QC measurements, filtering records and a filtered H5AD
scrna-scvi-umap	Filtered counts and a batch definition if needed	scVI or PCA representation, Leiden clusters and UMAP plots
scrna-cell-annotation	Clusters and a tissue-matched marker panel or HPA table	Candidate cell types, marker evidence and a review form
scrna-condition-de	Reviewed cell types, conditions and biological donor IDs	Donor-level pseudobulk counts and PyDESeq2 results

You can use one Skill or work through the stages. A sequencing report alone is enough for delivery review; downstream analysis also needs the count matrix.

A first request

Use scRNA