Conciseness

The body is dense with accurate, non-padded material (executable code, validator table, CLI commands, anti-patterns, pitfalls), but at ~790 lines inlined in SKILL.md, whole topics that belong in reference files (the validator type table, three separate auth provider setups, the Angular integration section, vector/pagination examples) inflate the token cost. Not 2 because it never explains basics Claude already knows and is well organized; not 4 because the sheer volume of inlinable reference material goes beyond 'minor instances of over-explanation'.

Actionability

Mostly executable, copy-paste-ready guidance: complete setup commands ("npx convex dev"), full schema/query/mutation/action definitions, React hook components, cron definitions, and storage functions covering the common cases. Not 5 because of minor gaps — the Stripe webhook leaves a "// Verify webhook signature here..." stub, and examples call functions never defined (api.emails.recordSent, api.documents.saveEmbedding, api.notifications.send).

Workflow Clarity

Setup is clearly sequenced (numbered "npx convex dev" steps, ordered sections from schema to functions to client), and the Problem/Solution pitfalls section offers some error recovery. But workflows involving database operations lack explicit validation checkpoints — deploy/import carry no verify step and schema changes have no validate-before-proceed loop — and the rubric caps database-operation skills without feedback loops at 3. Not 4 because the checkpoints that do exist (pitfalls) are detached from the workflows rather than embedded in them.

Progressive Disclosure

Section structure is clear and logically ordered, but there are no bundle files at all: all 790 lines live in SKILL.md with zero one-level-deep references, and content that clearly belongs in separate files (auth provider variants, Angular integration, validator API reference) is inlined. Matches anchor 3 ('content that should be separate is inline' with some structure); not 2 because structure is well organized, not minimal; not 4 because there is no reference splitting whatsoever.

Specificity

Lists seven concrete capability areas — "schema design, TypeScript functions, real-time subscriptions, auth, file storage, scheduling, and deployment" — which is broad, comprehensive coverage of the Convex surface. Not 5 because capabilities are phrased as domain nouns rather than concrete actions (anchor 5 examples use verbs like "Extract", "fill", "merge") and minor gaps exist (indexing, pagination, cron jobs); not 3 because it lists far more than the 1-2 concrete actions that anchor requires.

Completeness

The "what" is clear and specific ("Convex reactive backend expert: schema design, TypeScript functions, real-time subscriptions, auth, file storage, scheduling, and deployment"), but there is no "Use when..." clause or equivalent explicit trigger guidance, capping completeness at 3 per the judging guidelines. Not 4 because the "when" is entirely absent rather than merely under-specified.

Trigger Term Quality

Good keyword coverage: "Convex", "schema", "TypeScript", "real-time subscriptions", "auth", "file storage", "scheduling", "deployment" are terms users would naturally say. Not 5 because common natural variations are missing — "queries", "mutations", "actions", "cron jobs", "reactive database"/"BaaS" — which anchor 5 expects (synonyms and variations); not 3 because coverage goes well beyond a single relevant keyword.

Distinctiveness Conflict Risk

"Convex reactive backend expert" carves out a clear niche with a distinct, unambiguous trigger — a user needing this skill will say "Convex" — so conflict risk with other backend/BaaS skills is minimal. Stays at 5 because the domain noun anchors the description and none of the secondary terms would plausibly pull in the wrong skill.