The body is lean and delegates detail to references, but the Check/Fix/Explain/Code Review sections restate the same ARIA, focus-trap, and keyboard points in slightly different framings, so it could be tightened — not the score 3 'every token earns its place' anchor.

Gives concrete, specific guidance — exact attribute values (role='dialog', aria-modal='true', aria-labelledby) and exact behaviors (Tab cycles within modal, Escape closes, return focus) — which is actionable; per the instruction-only note, absence of code blocks is not penalized when guidance is this specific.

Criteria are listed as a flat Quick Reference checklist with no explicit sequenced steps or validate-then-retry feedback loop in the body, matching the score 2 'steps present but checkpoints missing or implicit' anchor rather than the score 3 explicit-checkpoint anchor.

Concise overview with a single, clearly signaled one-level-deep reference ('see references/rule.md'), and that bundle file exists as referenced, matching the well-organized one-level-deep anchor.

Lists multiple concrete actions — 'Check native semantics first, then inspect keyboard behavior, focus flow, accessible names, and screen-reader output' — matching the multi-action anchor rather than the single-domain score 2.

Explicitly answers both what ('Make modal dialogs keyboard accessible... inspect keyboard behavior, focus flow, accessible names') and when via a present 'Use when reviewing...' clause, satisfying the explicit-trigger requirement that would otherwise cap at 2.

Covers natural terms a user would say — 'modal dialogs', 'keyboard accessible', 'rendered HTML', 'interactive components' — with good variation; not merely technical jargon, so it is above the score 2 'missing common variations' anchor.

Scoped to a clear niche (modal dialog keyboard accessibility) with distinct triggers, unlikely to fire for unrelated skills; above the 'could still overlap' score 2 anchor.