scvi-tools

Deep generative models for single-cell omics; use when you need probabilistic batch correction (scVI), transfer learning, uncertainty-aware differential expression, or multimodal integration (totalVI/MultiVI).

Quality

86%

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

82%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.

A well-built skill body: a complete runnable example, clear capability catalog, and disciplined offloading of model/theory detail to real one-level-deep reference files. Weaknesses are mild — no explicit validation checkpoints in the training workflow and one orphaned bundle file (workflows.md) that the main file never points to.

Suggestions

Add explicit validation checkpoints to the example workflow, e.g. assert the counts layer exists and is non-negative before setup_anndata (adata.layers['counts'] is raw counts), and note how to check training convergence before extracting the latent representation.

Reference references/workflows.md from SKILL.md (e.g., in a 'Common workflows' section) so the bundle file is discoverable rather than orphaned.

Trim Implementation Details entries that restate well-known concepts (e.g., what variational inference / amortized inference does) and keep only scvi-tools-specific guidance; move version-sensitive pins to a dedicated compatibility note if they need to stay.

Dimension	Reasoning	Score
Conciseness	Largely efficient: dense When-to-Use and Key-Features lists plus a runnable example with no padding. Minor over-explanation in Implementation Details (re-stating that VAE-style variational inference learns a latent representation) and unpinned-to-a-section version numbers in Dependencies keep it below lean anchor 5.	4 / 5
Actionability	Fully executable, copy-paste-ready: a complete scVI workflow (setup_anndata with layer/batch/covariate keys, train, get_latent_representation, get_normalized_expression, Scanpy neighbors/UMAP/leiden, differential_expression with mode/delta), plus install commands and model save/load — covers the common batch-correction case end to end.	5 / 5
Workflow Clarity	The example is a clearly numbered 1–6 sequence with inline guard comments ("raw counts layer (not log-normalized)", batch_key/covariate registration), exceeding a bare step list. However, checkpoints are only implicit — no validation that the counts layer or batch key exist, no training-convergence check, and no error-recovery loop — so it stops short of anchor 5.	4 / 5
Progressive Disclosure	Good structure with well-signaled, one-level-deep references: per-modality model catalogs (references/models-*.md), differential-expression.md, and theoretical-foundations.md, all of which exist in the bundle. One organization gap: references/workflows.md exists in the bundle but is never referenced or navigated from SKILL.md.	4 / 5
	Total	17 / 20 Passed

Description

87%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.

A strong description that names the domain, lists several concrete capabilities with model identifiers, and provides an explicit 'use when' trigger clause. Coverage is nearly comprehensive but omits some natural synonyms (scRNA-seq, CITE-seq) and the specialized modalities the body actually supports.

Dimension	Reasoning	Score
Specificity	Lists several specific concrete actions with named models ("probabilistic batch correction (scVI), transfer learning, uncertainty-aware differential expression, or multimodal integration (totalVI/MultiVI)"). Not comprehensive: specialized modalities covered by the body (ATAC, spatial, doublet detection) are absent, so it falls just short of the anchor-5 example's full coverage.	4 / 5
Completeness	Explicitly answers both: what ("Deep generative models for single-cell omics") and when ("use when you need probabilistic batch correction (scVI), transfer learning, uncertainty-aware differential expression, or multimodal integration") with concrete trigger phrases, matching the anchor-5 example.	5 / 5
Trigger Term Quality	Good natural-phrase coverage: "single-cell", "batch correction", "transfer learning", "differential expression", "multimodal integration". A few common user terms/synonyms are missing, e.g. "scRNA-seq", "CITE-seq", "data integration", "cell annotation", so it is below the anchor-5 comprehensive-synonym example.	4 / 5
Distinctiveness Conflict Risk	Clear niche — probabilistic, model-based single-cell analysis — with distinct trigger capabilities and named models (scVI, totalVI, MultiVI) that distinguish it from generic Scanpy/pipeline skills; minimal conflict risk.	5 / 5
	Total	18 / 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 1 missing	Warning

	Total	14 / 16 Passed

Repository: aipoch/medical-research-skills
Path: scientific-skills/Data Analysis/scvi-tools/SKILL.md
Commit: 686e09d

Reviewed: 13 hours 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.