45+ scRNA-seq Skills — But Which Ones Should You Actually Install?
Single-cell RNA-seq analysis is one of the most skill-rich domains in the OpenClaw ecosystem. We found 45+ skills across three major repositories that handle some aspect of scRNA-seq. That's great for coverage, but terrible for decision-making.
This guide cuts through the noise. We tell you exactly which skill to install for each step of your workflow.
TL;DR — The Three You Need
If you install nothing else, install these:
| Skill | What it does | Install |
|---|---|---|
| scrna-orchestrator (ClawBio) | End-to-end pipeline: QC → clustering → annotation → report | clawhub install scrna-orchestrator |
| bio-single-cell-cell-annotation (FreedomIntelligence) | Automated cell type annotation with CellTypist/SingleR/Azimuth | clawhub install bio-single-cell-cell-annotation |
| scvelo (K-Dense) | RNA velocity, developmental trajectories, driver genes | clawhub install scvelo |
One all-rounder + two specialists. That covers 80% of scRNA-seq use cases.
The Full Workflow: Best Skill Per Step
1. Data Loading
Best pick: scanpy (K-Dense)
Handles 10X MTX, h5ad, h5, and raw count matrices. The foundational skill that most other scRNA-seq skills depend on.
clawhub install scanpyAlso consider: bio-single-cell-data-io (FreedomIntelligence) if you work with Seurat RDS files in R.
2. QC & Filtering
Best pick: scrna-orchestrator (ClawBio) — includes QC as part of its pipeline
If you want QC as a standalone step: scrna-qc (FreedomIntelligence) — MAD-based filtering with before/after plots and a qc_summary.json output.
3. Doublet Removal
Best pick: bio-single-cell-doublet-detection (FreedomIntelligence)
Supports three engines: Scrublet (Python), DoubletFinder (R), and scDblFinder (R). No API key needed.
4. Normalization, HVG, Dimensionality Reduction, Clustering
Best pick: scrna-orchestrator (ClawBio) — handles all of these in one pass
Does normalization → log1p → HVG selection → PCA (or scVI latent) → UMAP → Leiden clustering. Outputs a complete h5ad with all embeddings.
For scVI-based batch-aware embedding specifically: pair with scrna-embedding (ClawBio).
5. Cell Type Annotation
Best pick: bio-single-cell-cell-annotation (FreedomIntelligence)
Three annotation engines in one skill:
- CellTypist — deep learning, pre-trained on 100+ tissue models
- SingleR — reference-based correlation (R)
- Azimuth — Seurat's reference mapping
Also worth trying: universal-single-cell-annotator (FreedomIntelligence) — adds LLM-reasoning annotation on top of marker-based and CellTypist methods.
6. Batch Integration
Best pick: bio-single-cell-batch-integration (FreedomIntelligence)
Supports Harmony, scVI, Seurat CCA/RPCA, and fastMNN. Covers both Python and R workflows.
For scVI-only integration: scrna-embedding (ClawBio) — trains a local scVI model with --batch-key support.
7. Trajectory & Pseudotime
Best pick: scvelo (K-Dense)
RNA velocity from spliced/unspliced dynamics. Dynamical mode for accurate latent time estimation and driver gene identification.
For multi-method comparison: single-trajectory (FreedomIntelligence) — runs PAGA, Palantir, VIA, and scVelo in one skill via OmicVerse.
For Monocle3/Slingshot (R): bio-single-cell-trajectory-inference (FreedomIntelligence).
8. Cell-Cell Communication
Best pick: bio-single-cell-cell-communication (FreedomIntelligence)
CellChat + NicheNet + LIANA — three major ligand-receptor inference frameworks in one skill.
Also consider: single-cellphone-db for CellPhoneDB v5 specifically.
9. Marker Genes & DEG
Best pick: bio-single-cell-markers-annotation (FreedomIntelligence)
Wilcoxon, t-test, DESeq2 for cluster markers. Gene set scoring for functional annotation.
Also built into scrna-orchestrator (ClawBio) if you're using the pipeline approach.
10. Advanced / Specialized
| Task | Best Skill | Repo |
|---|---|---|
| scATAC-seq | bio-single-cell-scatac-analysis | FreedomIntelligence |
| CITE-seq / Multiome | bio-single-cell-multimodal-integration | FreedomIntelligence |
| Perturb-seq / CRISPR screens | bio-single-cell-perturb-seq | FreedomIntelligence |
| Splicing analysis | bio-single-cell-splicing | FreedomIntelligence |
| Lineage tracing | bio-single-cell-lineage-tracing | FreedomIntelligence |
| GRN inference | arboreto | K-Dense |
| Atlas queries (61M+ cells) | cellxgene-census | K-Dense |
| Spatial + scRNA integration | single-to-spatial-mapping | FreedomIntelligence |
| T-cell exhaustion profiling | tcell-exhaustion-analysis-agent | FreedomIntelligence |
Where Do These Skills Live?
| Repository | scRNA Skills | Strengths |
|---|---|---|
| K-Dense (15.7K stars) | 8 | Core tools: scanpy, scvelo, scvi-tools, cellxgene-census |
| FreedomIntelligence (1.5K stars) | 30+ | Most comprehensive: modular bio-single-cell-* suite + specialist agents |
| ClawBio (471 stars) | 2 | Best end-to-end: scrna-orchestrator + scrna-embedding with reproducibility bundles |
None of these skills require API keys. All run locally with standard Python/R packages.
Our Recommendation
Start with scrna-orchestrator from ClawBio. It handles the entire standard workflow in one command and generates a reproducibility bundle (commands.sh, environment.yml, SHA-256 checksums). When you need to go deeper — cell type annotation, RNA velocity, or cell communication — add the specialist skills from K-Dense and FreedomIntelligence.
Think of it as: ClawBio for the pipeline, K-Dense for the core tools, FreedomIntelligence for the specialists.
Last updated: March 21, 2026. All skills surveyed are actively maintained and MIT-licensed.