Lean and dense with no concept padding — it never explains what streams are and assumes Claude's competence, with every line being a rule, table row, or executable snippet that earns its place.

Provides fully executable code examples and concrete API tables (e.g. 'filter(...).findAny()', 'anyMatch / noneMatch / allMatch', 'Collectors.joining', Java-version-gated 'toList()' and 'teeing'), all copy-paste ready.

A clearly sequenced Core Workflow (steps 0–7) with an explicit validation checkpoint at step 7 — 'Verify changed branches for empty inputs, one element, duplicates, nulls, ordering, parallel-safety, and baseline compatibility. Run the marker scan... fix hits, and re-scan' — giving a fix-and-revalidate feedback loop.

Body is an overview with a File/Purpose table signaling three one-level-deep references (hard-stops.md, stream-examples.md, java-stream-api.md), all of which exist as real bundle files, with content appropriately split out of SKILL.md.

Lists multiple concrete actions — 'Review Java stream performance advice', 'clean up mutation and write or refactor Java Stream and Collector code', and named antipatterns like 'materializing just to inspect, sorting before min/max, counting for existence' — rather than vague language.

Explicitly answers both what (review/clean up/refactor/avoid antipatterns) and when via an explicit 'Use whenever writing, reviewing, or refactoring Java code that uses Java streams...' trigger clause.

Covers natural terms users would say — 'Java streams, collectors, stream pipelines, grouping, joining strings, first/any element lookup, sorting, limiting, distinct values, primitive totals, Optional values in streams, or parallel streams' — including the findFirst/findAny phrasing reviewers actually use.

Occupies a clear niche (Java stream/collector performance and antipattern review) with stream-specific triggers, making it unlikely to fire for unrelated skills; written in third person with no voice penalty.