Conciseness

The 46-line body is all functional tables and terse imperative steps ("Never transform MCP results in a shell or interpreter"), with zero explanation of concepts Claude already knows and no padding. Matches anchor 5 — every token earns its place; the one-line 'When to Use' section is negligible duplication of the frontmatter.

Actionability

Guidance is directive and specific ("Confirm scope, period, currency, and breakdown"; error table actions like "Retry once; then stop with the trace ID"), but the body contains no directly executable tool-call detail — no inline query_costs parameters, and "Resource Graph" is named without a tool name — leaving the executable specifics one reference away. Anchor 4 (mostly executable, minor gaps) fits better than anchor 5 (nothing in the body is copy-paste ready) and better than anchor 3 (the guidance is concrete, not vague or pseudocode).

Workflow Clarity

A clear 5-step sequence with an upfront confirmation checkpoint (step 1), an incomplete-data labeling checkpoint (step 4), and an Error Handling table with recovery actions. It sits at anchor 4 rather than 5 because the steps read as policy directives and the checkpoints are less explicit/conditional than the anchor-5 validate-fix-retry loop; this is a read-only query skill, so the destructive/batch cap at 3 does not apply.

Progressive Disclosure

The Quick Reference table is a clear intent-to-workflow router to real files (all five referenced paths exist in the bundle) and the body stays lean. However, references run two levels deep from SKILL.md (cost-investigation.md links onward to aks-cost-analysis.md which links to aks-anomaly.md/aks-enablement.md; cost-query/workflow.md links to four siblings), deviating from anchor 5's 'one-level-deep references' — anchor 4's 'good structure, minor organization gaps' is the best fit.

Specificity

"Analyze actual Azure spend, bill changes, and AKS or AI costs" names the domain and several concrete capability areas (spend analysis, bill changes, AKS allocation, Cosmos DB / Foundry costs), matching anchor 4's 'several specific actions; minor gaps'. It falls short of anchor 5 because the action vocabulary is essentially one verb ('Analyze') over multiple objects rather than multiple distinct concrete actions, and ahead of anchor 3 because coverage spans more than 1-2 actions.

Completeness

It explicitly answers both 'what' ("Analyze actual Azure spend, bill changes, and AKS or AI costs") and 'when' (explicit "WHEN:" clause with concrete trigger phrases, plus a "DO NOT USE FOR" exclusion), which matches anchor 5. Anchor 4 ('when' could be more explicit) does not apply since the trigger guidance is labeled and concrete.

Trigger Term Quality

Eleven natural trigger phrases — "Azure cost breakdown", "most expensive resources", "why did my bill increase", "cost spike", "unexpected charge", "Cosmos DB charges", "AKS cluster cost", "idle AKS capacity" — give comprehensive coverage with synonyms (cost/charge/bill/spend) and question-form phrasing users would actually say. This matches anchor 5 rather than anchor 4 ('a few natural terms missing') since the list spans the skill's entire scope across spend, spikes, per-service, AKS, and AI.

Distinctiveness Conflict Risk

Azure-specific terms (AKS, Cosmos DB, Foundry) establish a clear niche, and "DO NOT USE FOR: forecasts, pricing estimates, rightsizing, commitments, or budgets" explicitly prevents triggering for adjacent cost skills. Minimal conflict risk, matching anchor 5.