Conciseness

The same 8-phase pipeline is restated three times — the 'Workflow Overview' ASCII diagram, the 'Phase Summaries' section, and the 'Core Capabilities' bullets — which is noticeable redundancy, though much of the domain-specific material (QC cutoffs, Moran's I thresholds, tool parameters) is genuinely additive. Not 4: the triplication and ~100 lines of HuBMAP tool detail exceed 'minor instances of over-explanation'.

Actionability

Concrete, executable guidance throughout: specific thresholds ('min 200 genes, 500 UMI, <20% MT', 'FDR < 0.05', 'Moran's I > 0.3'), tool calls with parameters and documented return shapes (e.g., HuBMAP_search_samples with organ/sample_category/limit), and runnable scanpy snippets. Not 5: the full pipeline code is delegated to code_examples.md, which is not present in the bundle, leaving the core analysis loop described rather than copy-paste ready.

Workflow Clarity

The 8-phase sequence is clearly ordered with checkpoints — QC gates in Phase 1, alignment verification, FDR filtering in Phase 4, marker-gene validation in Phase 6, and a Quantified Minimums checklist plus T1-T4 evidence grading for interpretation. Not 5: there are no error-recovery feedback loops (e.g., what to do when QC filters remove too many spots or alignment fails); checkpoints state 'verify' without recovery steps.

Progressive Disclosure

References are clearly signaled and one level deep ('[code_examples.md](code_examples.md)', '[report_template.md](report_template.md)'), but neither file exists in the bundle — the links are broken — and ~100 lines of HuBMAP tool documentation that clearly belong in a separate reference file are inlined in SKILL.md. Not 4: broken reference targets plus inlined reference-grade content are more than 'minor organization gaps'.

Specificity

Lists four concrete, distinct actions — 'Maps gene expression to tissue architecture, identifies spatially variable genes (SVGs), tissue-domain segmentation, and cell-cell interaction inference' — plus a named platform list (Visium, MERFISH, seqFISH, Slide-seq), giving comprehensive coverage of the skill's capabilities. Not 4: no meaningful capability of the domain is left out, matching the 'multiple specific concrete actions; comprehensive coverage' anchor.

Completeness

Explicitly answers both: what it does ('Maps gene expression to tissue architecture, identifies... SVGs, tissue-domain segmentation, and cell-cell interaction inference') and when to use it ('Use for spatial gene-expression questions, tissue architecture analysis, and SVG identification') with concrete trigger phrases. Not 4: the 'when' clause is explicit and specific rather than merely adequate.

Trigger Term Quality

Platform names (Visium, MERFISH, seqFISH, Slide-seq) and phrases like 'spatial gene-expression questions' and 'tissue architecture analysis' are natural user terms with good coverage. Not 5: common variations users would say — 'spatial data', 'spatial omics', 'spatial domains', file extensions like .h5ad — are absent.

Distinctiveness Conflict Risk

Clear niche (spatially-resolved transcriptomics) with distinct platform-specific triggers (Visium, MERFISH, seqFISH, Slide-seq) unlikely to fire for other skills. Not 4: the platform names and SVG/tissue-domain terminology create minimal overlap risk even against a general single-cell analysis skill.