scvi-tools
Deep generative models for single-cell omics. Use when you need probabilistic batch correction (scVI), transfer learning, differential expression with uncertainty, or multi-modal integration (TOTALVI, MultiVI). Best for advanced modeling, batch effects, multimodal data. For standard analysis pipelines use scanpy.
92
86%
Does it follow best practices?
Impact
97%
1.14xAverage score across 6 eval scenarios
Passed
No known issues
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is an excellent skill description that clearly defines its domain (deep generative models for single-cell omics), lists specific concrete capabilities with tool names, provides explicit 'Use when' triggers, and even includes a boundary condition distinguishing it from the scanpy skill. The description is concise yet comprehensive, using appropriate domain-specific terminology that users would naturally employ.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: probabilistic batch correction (scVI), transfer learning, differential expression with uncertainty, and multi-modal integration (TOTALVI, MultiVI). Also distinguishes from standard analysis pipelines. | 3 / 3 |
Completeness | Clearly answers both 'what' (deep generative models for single-cell omics with specific capabilities listed) and 'when' (explicit 'Use when' clause with specific triggers, plus 'Best for' guidance and a boundary condition directing users to scanpy for standard pipelines). | 3 / 3 |
Trigger Term Quality | Includes strong natural keywords users in this domain would use: 'scVI', 'TOTALVI', 'MultiVI', 'batch correction', 'transfer learning', 'differential expression', 'multi-modal integration', 'single-cell omics', 'multimodal data', 'batch effects'. Good coverage of both tool names and conceptual terms. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive with a clear niche in deep generative modeling for single-cell omics. Explicitly differentiates from scanpy for standard analysis, reducing conflict risk. The specific tool names (scVI, TOTALVI, MultiVI) serve as unique identifiers. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
72%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured skill with strong actionability and excellent progressive disclosure, effectively serving as a hub document pointing to detailed references. Its main weaknesses are some verbosity (explaining concepts Claude already knows, like variational inference theory) and missing validation/error-recovery steps in the workflow. The code examples are high quality and directly executable.
Suggestions
Remove or drastically shorten the 'Theoretical Foundations' section — Claude already understands VAEs, variational inference, and amortized inference. A single line linking to the reference file would suffice.
Add validation checkpoints to the workflow: e.g., check training convergence via `model.history`, verify latent representation quality with UMAP visualization, and mention common failure modes (e.g., ensuring raw counts, handling GPU memory errors).
Trim the 'When to Use This Skill' section — the description/frontmatter already covers when to use it, and listing every possible modality is redundant with the 'Core Capabilities' section that follows.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill includes some unnecessary content like the 'Theoretical Foundations' section (Claude knows what VAEs and variational inference are) and the 'When to Use This Skill' section is overly exhaustive. The overview paragraph explaining what scvi-tools is built on is also redundant. However, the code examples and model listings are reasonably efficient. | 2 / 3 |
Actionability | The skill provides fully executable, copy-paste ready code examples for the typical workflow, differential expression, model persistence, and batch correction. The code includes specific function calls, parameter names, and realistic values that Claude can directly use. | 3 / 3 |
Workflow Clarity | The typical workflow is clearly numbered with 6 steps and includes concrete code, which is good. However, there are no validation checkpoints or error recovery steps — no guidance on checking training convergence, validating latent representations, or handling common failures like CUDA errors or data format issues. | 2 / 3 |
Progressive Disclosure | Excellent progressive disclosure structure: the main file provides a clear overview with concise descriptions of each model category, then points to specific reference files (references/models-scrna-seq.md, references/differential-expression.md, etc.) with well-signaled one-level-deep references. Content is appropriately split between overview and detailed references. | 3 / 3 |
Total | 10 / 12 Passed |
Validation
90%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 10 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
metadata_version | 'metadata.version' is missing | Warning |
Total | 10 / 11 Passed | |
- Repository
- K-Dense-AI/claude-scientific-skills
- Commit
1420470
- Reviewed
Table of Contents
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.