Conciseness

The body is dense policy guidance that mostly adds non-obvious, project-specific rules ('Use scDblFinder... run it per batch', 'Avoid whole-object copies... use views, backed mode') rather than explaining concepts Claude already knows. It is not level 5 because several workflow bullets (marker-panel selection, memory practices, plot legibility) duplicate material in references/qc-annotation-umap-heuristics.md and could be trimmed to pointers.

Actionability

Mostly actionable: one fully executable command ('python plugins/ngs-analysis/scripts/run_scrnaseq_post_count_qc.py --input-dir path/to/scrna_bundle') plus concrete tool-level directives (scDblFinder per batch, passes_QC column, MapMyCells/cell_type_mapper, scVI vs PCA decision criteria). Not level 5 because the core workflow steps have no example code or command, and the runner path referenced is not present in the skill's bundle.

Workflow Clarity

A clearly sequenced 7-step workflow (thresholds -> QC -> latent space -> annotation -> marker panel -> UMAP -> scaling) with most checkpoints present: threshold-justification plots, per-filter QC summary counts, cluster inspection before downstream steps, provenance manifests, and stage checkpoints. Not level 5 because there is no explicit validate -> fix -> retry loop (e.g., what to do when a threshold plot or cluster inspection fails review); verification exists but error-recovery feedback loops are implicit.

Progressive Disclosure

Good structure: a 'Start Here' pointer and a Resources section clearly signal the single one-level-deep reference (references/qc-annotation-umap-heuristics.md, verified to exist, with no nested references inside it), and detailed heuristics are appropriately split out. Not level 5 because the split is not clean — several topics appear both inline in SKILL.md and in the reference file — and the runner script path 'plugins/ngs-analysis/scripts/run_scrnaseq_post_count_qc.py' is not part of the bundle.

Specificity

Lists multiple concrete actions with comprehensive coverage: 'Process, quality-control, annotate, and visualize single-cell or single-nucleus RNA-seq datasets', 'run required scDblFinder-based doublet and ambient-RNA filtering', 'annotate cells with matched references or marker-based fallbacks', 'generate global and per-group UMAP visualizations'. This matches the level-5 anchor (multiple specific concrete actions, comprehensive); it is not level 4 because coverage extends beyond minor gaps into named methods and fallback strategies.

Completeness

Clearly answers both: the 'what' is 'Process, quality-control, annotate, and visualize... datasets across tissues and species', and the 'when' is explicit with concrete trigger phrases ('Use when... build, adapt, or review a general scRNA-seq QC pipeline; choose... filters; run... doublet... filtering; annotate...; or generate... UMAP visualizations'). This matches the level-5 anchor exactly; a missing or weak 'when' clause would be required for a lower score.

Trigger Term Quality

Good keyword coverage with natural domain terms users would say: 'single-cell or single-nucleus RNA-seq', 'scRNA-seq', 'QC pipeline', 'doublet', 'ambient-RNA', 'UMAP', 'annotate'. Not level 5 because a few natural synonyms are missing (e.g., 'cell typing', 'label transfer', 'batch integration', 'AnnData/Scanpy/Seurat'); not level 3 because the included terms are varied and cover both the coarse and specific phrasings.

Distinctiveness Conflict Risk

Clear niche with distinct triggers: scRNA-seq/snRNA-seq QC, scDblFinder doublet calling, matched-reference annotation, global/per-group UMAP. Minimal overlap risk with adjacent bioinformatics skills; it is well above the level-4 'minor overlap risk with closely related skills' anchor.