single-cell-ml

FGCZ Skills

Shared Agent Skills for bioinformatics workflows at the Functional Genomics Center Zurich (FGCZ).

This repository contains reusable skills that extend AI agents (Claude Code, Cursor, VS Code, etc.) with FGCZ-specific knowledge, workflows, and infrastructure integration.

Another example of biology related skill list can be found in the bioSkills repo.

What are Agent Skills?

Agent Skills are a lightweight, open format for extending AI agent capabilities with specialized knowledge and workflows. Originally developed by Anthropic and released as an open standard, skills are now supported by multiple AI development tools.

Key benefits:

• Domain expertise - Package specialized bioinformatics knowledge
• Repeatable workflows - Convert multi-step tasks into consistent, auditable workflows
• Portable - Works across Claude Code, Cursor, VS Code, and other skills-compatible agents
• Version controlled - Skills are just files, easy to version and share

Quick Start

Using with Claude Code — Plugin Marketplace (Recommended)

This repo is a Claude Code plugin marketplace. The marketplace flow keeps every team member on the same version, supports auto-update, and avoids manual path config.

1. Add the marketplace (one-time, per machine):

   claude /plugin marketplace add fgcz/skills
   

2. Install the plugins you need (cherry-pick by role):

   # Single-cell analysts:
   claude /plugin install single-cell-spatial-general@fgcz-skills
   claude /plugin install single-cell-ml@fgcz-skills
   
   # Pipelines & demux:
   claude /plugin install sequencing-pipelines@fgcz-skills
   
   # B-Fabric / LIMS:
   claude /plugin install bfabric-lims@fgcz-skills
   
   # Infrastructure / SUSHI app dev:
   claude /plugin install fgcz-infrastructure@fgcz-skills
   
   # Metabolomics:
   claude /plugin install metabolomics-data-analysis@fgcz-skills
   

3. Enable auto-update so new releases land without manual pulls. Add to ~/.claude/settings.json:

   {
     "extraKnownMarketplaces": {
       "fgcz-skills": {
         "source": { "source": "github", "repo": "fgcz/skills" },
         "autoUpdate": true
       }
     }
   }
   

4. Verify: claude /plugin list should show the installed plugins with their version.

Skills auto-activate from keyword matching in the description field — just mention relevant terms (e.g. "Seurat", "CellRanger", "B-Fabric") in your prompt.

Legacy: clone-and-symlink (no auto-update)

If you cannot use the marketplace (offline cluster, custom setup), clone the repo and point Claude Code at the relevant plugin's skills/ directory. As of v1.3.0, skills live under per-plugin subdirectories (one skills/ dir per plugin):

git clone [email protected]:fgcz/skills.git ~/.claude/fgcz-skills

Add the plugins you need to ~/.claude/settings.json:

{
  "skills": {
    "paths": [
      "~/.claude/fgcz-skills/single-cell-spatial-general/skills",
      "~/.claude/fgcz-skills/fgcz-infrastructure/skills",
      "~/.claude/fgcz-skills/sequencing-pipelines/skills"
    ]
  }
}

This path skips the marketplace's version pinning and auto-update — you're responsible for pulling the repo manually.

Using with Other Agents

Agent Skills are compatible with any skills-compatible agent. See the integration guide for details.

Available Skills

Single-cell & Spatial Transcriptomics (single-cell-spatial-general plugin)

Skill	Description
seurat-analysis	Seurat v5 workflows: QC, normalization, clustering, DEGs, integration
cellranger-fgcz	10x CellRanger processing: HTO, OCM, CITE-seq, VDJ, multi config
cytotrace2-analysis	Cellular potency prediction and differentiation states
screpertoire-analysis	TCR/BCR repertoire analysis with scRepertoire
rctd-annotation	Spatial cell type deconvolution with RCTD
split-purification	Xenium/VisiumHD contamination removal with SPLIT
scevan-analysis	CNV-based malignant cell identification
scomatic-analysis	Somatic mutation detection in scRNA-seq
cellchat-analysis	Cell-cell communication inference + per-condition merge/compare
slingshot-trajectory	Trajectory inference and pseudotime on Seurat v5 embeddings
xenium-ccf-registration	Xenium → Allen Brain CCFv3 alignment via STalign LDDMM
clusterprofiler-pathways	GO/KEGG/MSigDB/Reactome/GSEA pathway enrichment on DEGs
insitucnv-analysis	Spatial CNV inference (Xenium, MERFISH) via infercnvpy

Deep Learning & Advanced Methods (single-cell-ml plugin)