Conciseness

The prose is mostly efficient and assumes Claude's competence, but the "Output Contract" section nearly duplicates the return guidance in "Creating Or Updating The Layer" (path or installed link, source coverage, caveats, refresh offer, iteration guidance), and "For ordinary data work, return to the Data Analytics `index` workflow" appears twice. This is real tightening work, matching anchor 3 rather than anchor 4's 'minor instances that could be trimmed'.

Actionability

Concrete elements exist (the `$CODEX_HOME/skills/<area>-semantic-layer` destination, the destination decision table, named reference paths), but key steps stay abstract: "Build and validate the same canonical semantic-layer skill package" and "Validate the written skill and verify it is discoverable" never specify how, and the bundle's `scripts/validate_data_context_contract.py` is never invoked. That is concrete guidance with missing executable detail, i.e. anchor 3, not anchor 4's mostly-executable guidance.

Workflow Clarity

The sequence is coherent across sections (routing → source inventory → crawl → create → validate → offer refresh → report), but validation checkpoints are named without commands or error-recovery loops — no 'if validation fails, do X' path — and the existing validation script is unreferenced. Steps present with checkpoint gaps matches anchor 3, below anchor 4's 'most checkpoints present'.

Progressive Disclosure

References are one level deep and body-referenced paths (`references/semantic-layer/skill-template.md`, `references/semantic-layer/weekly-polling-automation.md`) resolve to real substantive files, but four of nine bundle files — all three scripts, `references/automation.md`, `references/semantic-layer/connector-playbook.md`, and `references/semantic-layer/source-intake.md` — are never signaled from the body, so navigation to them depends on guessing. Structure exists but references are partly buried, matching anchor 3 rather than anchor 4.

Specificity

"Create, update, inspect, or repair Data Analytics semantic layers" lists several specific lifecycle actions with comprehensive coverage of that lifecycle, but the verbs all operate on one object and never name concrete layer contents (e.g. metric definitions, source-of-truth dashboards), matching anchor 4's 'minor gaps in coverage' rather than anchor 5's multiple distinct concrete actions.

Completeness

The description explicitly answers both: what ("Create, update, inspect, or repair Data Analytics semantic layers") and when ("Use when the user asks to save data context or create a semantic layer that future Data Analytics work can inspect and cite") with concrete trigger phrases, matching the anchor 5 example structure exactly; it is not a 4 because the 'when' clause is fully explicit rather than merely present.

Trigger Term Quality

Natural user phrases like "save data context" and "create a semantic layer" are present, but common variations users would actually say — "data dictionary", "canonical metric definitions", "document data sources" — are missing, fitting anchor 4's 'good keyword coverage; a few natural terms missing' rather than anchor 5's synonym-and-extension completeness.

Distinctiveness Conflict Risk

"Data Analytics semantic layers" and "save data context" carve a clear niche with distinct triggers, but the broad "Data Analytics" domain term leaves minor overlap risk with general data-analysis skills, matching anchor 4 rather than anchor 5's minimal-conflict niche.