scvi-tools

Probabilistic single-cell RNA-seq with scvi-tools — scVI for a batch-corrected latent space, scANVI for semi-supervised label transfer, and Bayesian differential expression. Reach for this skill to integrate scRNA-seq batches, embed cells for clustering, transfer annotations from a reference onto a query, or score differentially expressed genes per cluster. For spatial deconvolution / mapping use the cell2location, DestVI, or Tangram methods instead.

Quality

91%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

88%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is a lean, highly actionable reference built around real failure modes with explicit validation and error-recovery guidance. Its main weaknesses are duplication between the Gotchas and Troubleshooting tables and a dangling kernel.py reference with no bundle present.

Suggestions

Merge the overlapping rows of the Gotchas and Troubleshooting tables (lfc_mean/IORegistryError/use_gpu appear in both with near-identical fixes) into a single symptom→cause→fix table.

Ship kernel.py in the skill bundle (e.g. scripts/kernel.py) or inline the two helper definitions, since the exec(open("scvi-tools/kernel.py")) path resolves to no file present in the skill directory.

Trim the 14-column vanilla DE listing to just the key absence (no lfc_*/proba_de) and link the full schema to a reference file, keeping the change-mode guidance inline.

Dimension	Reasoning	Score
Conciseness	The body is dense and assumes competence — it documents non-obvious failure modes (vanilla DE default, ArrowStringArray write error, removed use_gpu kwarg) without explaining known concepts. However the Gotchas and Troubleshooting tables repeat the same three errors with near-identical fixes, and the 14-column vanilla DE listing is heavy. Not 5 because this duplication could be trimmed; not 3 because over-explanation is minor relative to the overall lean content.	4 / 5
Actionability	Copy-paste-ready scVI, scANVI, and DE recipes with exact kwargs (setup_anndata(layer="counts", batch_key="batch"), accelerator="gpu", devices=1) and a version-pinned Modal script with concrete CLI invocation. Not 4 because the common cases are covered by fully executable code; the only placeholder ("see the recipe above") is an explicitly justified cross-reference.	5 / 5
Workflow Clarity	Clear sequence (load helpers → prepare/validate counts → train → embed → DE → write) with an explicit validation checkpoint (prepare_scvi_counts, provenance verification before training) and error→fix feedback loops (Troubleshooting table plus the NameError recovery hint). Not 4 because checkpoints are explicit, not merely implied.	5 / 5
Progressive Disclosure	Well-sectioned single file with clear headers, tables, and a Next pointer. However the body references "scvi-tools/kernel.py" and a "remote-compute-modal" skill while no bundle files (references/, scripts/, assets/) exist in this skill's directory, so the primary referenced path cannot be resolved. Not 5 because the kernel.py reference is a navigation gap; not 3 because the inline content is appropriately organized for a single-file skill.	4 / 5
	Total	18 / 20 Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is specific, third-person, and explicitly covers both what the skill does and when to reach for it, with a boundary clause steering spatial use cases elsewhere. Only weakness is modest synonym coverage in trigger terms.

Dimension	Reasoning	Score
Specificity	"scVI for a batch-corrected latent space, scANVI for semi-supervised label transfer, and Bayesian differential expression" lists multiple specific concrete actions with comprehensive coverage of the tool's capabilities. Not 4 because coverage of the skill's three core actions is complete rather than having minor gaps.	5 / 5
Completeness	Explicitly answers what ("scVI for a batch-corrected latent space, scANVI for semi-supervised label transfer, and Bayesian differential expression") and when ("Reach for this skill to integrate scRNA-seq batches, embed cells...") with concrete trigger phrases. Not 4 because both what and when are explicit and specific rather than merely present.	5 / 5
Trigger Term Quality	"integrate scRNA-seq batches, embed cells for clustering, transfer annotations from a reference onto a query, score differentially expressed genes" are natural user phrases, but common synonyms like "batch correction", "cell type annotation", or file extensions (.h5ad) are missing. Not 5 because keyword coverage is good rather than comprehensive; not 3 because several genuinely natural trigger phrases are present.	4 / 5
Distinctiveness Conflict Risk	Names a specific package (scvi-tools) and models (scVI/scANVI) in a clear niche, and actively reduces conflict risk by redirecting spatial work to "cell2location, DestVI, or Tangram methods instead". Not 4 because the redirect makes overlap with adjacent single-cell skills minimal.	5 / 5
	Total	19 / 20 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: aipoch/open-science
Path: resources/skills/scvi-tools/SKILL.md
Commit: bf35648

Reviewed: 8 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.