Conciseness

The body is code-dominant with terse comments and avoids explaining concepts Claude already knows (no primer on what RNA-seq or UMAP is). It is not 5 because the three overlapping advisory sections — 'Key Parameters to Adjust', 'Common Pitfalls and Best Practices', and 'Tips for Effective Analysis' — repeat each other (multi-resolution clustering advice and the use_raw note each appear two to three times) and could be consolidated.

Actionability

Fully executable, copy-paste-ready code throughout every workflow step, plus concrete CLI invocations with flags ('python scripts/qc_analysis.py input.h5ad --output filtered.h5ad --mt-threshold 5 --min-genes 200 --min-cells 3') and a copy-and-customize template command. Specific examples cover the common cases (10X/h5ad/CSV loading, Leiden clustering, marker identification).

Workflow Clarity

A clearly numbered 1-7 sequence (QC → normalization → dimensionality reduction → clustering → markers → annotation → save) with real checkpoints: QC violin plots before thresholding, the PCA elbow plot ('Check elbow plot'), trying multiple resolutions, and 'Validate biologically: Check marker genes match expected cell types'. It is not 5 because the destructive filtering step (filter_cells/filter_genes/MT% subsetting) lacks an explicit post-filter verification step (e.g., re-checking adata shape or re-plotting after filtering); it is above 3 because validation checkpoints are present throughout, not missing.

Progressive Disclosure

The bundle is well-structured and verified: all referenced files exist (scripts/qc_analysis.py, three references/*.md, assets/analysis_template.py), references are one level deep with no nested pointers, and each is clearly signaled with purpose and when-to-read guidance ('Read this reference when performing a complete analysis from scratch'). It is not 5 because the ~200-line inline standard workflow substantially duplicates references/standard_workflow.md, so the split between overview and detail is imperfect.

Specificity

The description lists multiple specific concrete actions — 'QC, normalization, dimensionality reduction (PCA/UMAP/t-SNE), clustering, differential expression, and visualization' — giving comprehensive coverage of the pipeline's capabilities. It is not score 4 because there is no meaningful gap in the coverage of the skill's core actions.

Completeness

It explicitly answers both what ('Standard single-cell RNA-seq analysis pipeline... QC, normalization... visualization') and when ('Use for QC...', 'Best for exploratory scRNA-seq analysis with established workflows'), with concrete trigger phrases plus negative boundary guidance ('For deep learning models use scvi-tools; for data format questions use anndata'). This clearly matches the top anchor.

Trigger Term Quality

Good natural keyword coverage including 'single-cell RNA-seq', 'scRNA-seq', 'QC', 'clustering', 'UMAP', 'PCA', and 'differential expression', which users would naturally say. It is not 5 because it omits file extensions (.h5ad, 10X) and common request phrases like 'marker genes' or 'cell type annotation' that the body covers.

Distinctiveness Conflict Risk

It carves out a clear niche (exploratory scRNA-seq analysis) and explicitly disambiguates against the two most confusable adjacent skills (scvi-tools for deep learning, anndata for data format questions), minimizing wrong-skill triggering. It is not 4 because the overlap risk is addressed head-on rather than merely minor.