Conciseness

The body is dense and imperative with no explanation of concepts Claude already knows ('Static review guesses; the deployment KNOWS'). Matches anchor 4 rather than 5 because the Rules section restates content already in the Workflow — the severity mapping ('limit-hit / permanent-OCC-failure = high; threshold / retried = med') duplicates step 4, and 'Stay in lane' duplicates step 6. Not 3 since the duplication is minor and everything else earns its tokens.

Actionability

Gives exact MCP tool names ('status', 'insights', 'tables', 'functionSpec'), concrete anti-patterns and fixes ('.collect() / unindexed .filter()', '.withIndex, .take(n), or .paginate', '@convex-dev/sharded-counter'), a sample evidence line, and an explicit fallback. Matches anchor 4 — mostly executable with minor gaps; not 5 because there is no copy-paste-ready example of an emitted finding against the referenced schema.

Workflow Clarity

Six clearly sequenced steps (GUARD → GATHER → ROOT-CAUSE → EMIT → REPORT → scope) with explicit validation and feedback loops: step 2's fallback ('if it returns nothing or is unavailable, say so and fall back... do NOT invent findings') and step 5's gate plus re-verification ('apply only on confirmation, then re-run `insights` after traffic to verify the trend'). Matches anchor 5; the per-event-kind root-cause mapping also functions as a checklist.

Progressive Disclosure

Under 50 lines with well-organized Workflow/Rules sections and appropriately inlined content, but the body twice directs emission 'per specs/finding.schema.json' and that file is not present in the bundle (no references/, scripts/, or assets/ exist), so the core output contract is unreadable. Anchor 4 (good structure, minor organization gap) rather than 5.

Specificity

Lists three concrete actions — 'Read the Convex deployment's 72h insights', 'root-cause each event in code', 'report evidence-backed perf/cost findings with fixes' — with parenthetical detail ('read limits, OCC contention'). Matches anchor 4 rather than 5 because coverage has minor gaps: log-failure triage and table-stat analysis from the body are not mentioned.

Completeness

The 'what' is clear (read insights → root-cause in code → report evidence-backed findings with fixes), but there is no 'Use when...' clause or equivalent explicit trigger guidance, which caps completeness at 3 per the judging guidelines. Not 4 because 'when' is entirely absent rather than merely implicit-but-present.

Trigger Term Quality

Contains genuinely natural domain terms a Convex user would say — 'Convex', 'deployment', 'read limits', 'OCC contention', 'perf/cost' — but misses common variations like 'slow queries', 'production health', or 'database performance'. Good coverage with a few natural terms missing fits anchor 4, not the comprehensive synonym coverage of anchor 5.

Distinctiveness Conflict Risk

Phrases like 'the Convex deployment's 72h insights' and 'OCC contention' carve out a clear niche with distinct triggers; virtually no other skill would fire on these terms. Matches anchor 5 (clear niche, minimal conflict risk) rather than 4, which requires some overlap with closely related skills.