The body assumes Claude's competence with no concept padding, but at ~320 lines it repeats scope-resolution guidance across Prerequisites, Step 1, 1.1, and 2.2 and could be tightened — matching the 'mostly efficient but could be tightened' anchor rather than level 3.

Fully executable guidance throughout — exact `node scripts/*.mjs` commands, jq filters, bash, and verbatim prompts to print — copy-paste ready, matching the level-3 anchor.

Clear numbered pipeline (Collect → Gate → Verify → Render) with explicit validation checkpoints ('Stop on blockers', verify-and-regen, regenPlan rerun) and feedback loops for error recovery, matching the level-3 anchor.

Well-signaled one-level-deep references to real files (doctrine.md, data-collection.md, candidates.md, etc.) with content appropriately split across scripts/ and references/, but `lib/queries.mjs` is referenced and does not exist, a broken navigation pointer that drops it below level 3.

Names multiple concrete actions — 'Collect Vercel metrics, usage, project config, and code scan results first; investigate only metric-backed candidates; produce ranked recommendations grounded in verified files and version-aware Vercel/framework docs' — comprehensive and specific, matching the level-3 anchor.

Clearly answers both 'what' (collect/investigate/produce ranked recommendations) and 'when' via an explicit 'Trigger for ...' clause, satisfying the level-3 anchor for both what AND when.

Explicit trigger clause covers natural user terms: 'Trigger for Vercel bill reduction, slow or expensive routes, caching opportunities, Function Invocations, Build Minutes, Fast Data Transfer, Core Web Vitals, Bot Management, Fluid compute, or cost breakdown requests.'

Scoped to a clear niche — Vercel cost/performance optimization on deployed projects with named frameworks — with distinct triggers unlikely to conflict with other skills.