Conciseness

The body is lean (~50 lines), never explains what design patterns are, and focuses on project-specific judgment rules ('模式选择要优先服从模块边界 + 扩展点稳定性', '如果只是 2 到 3 个稳定分支...先别急着上复杂模式'). Minor over-explanation appears in 高信号规则, which partially restates textbook knowledge Claude already has ('工厂模式适合如何创建，策略模式适合如何选择一种行为'). Not 5 because of that minor textbook restatement; not 3 because there is no real padding.

Actionability

The 快速指导 section gives a genuine decision procedure (classify problem type → check module boundaries → avoid over-engineering small branches → check existing implementations), which is concrete for an instruction-only skill. However, all three detail files it points to ('reference/1-core-patterns.md', 'reference/2-selection-guidelines.md', 'reference/3-project-examples.md') do not exist in the bundle — the primary actionable pointers are dangling. Not 4 because missing key detail files leave the guidance incomplete; not 2 because the inline decision tree itself is specific and executable.

Workflow Clarity

快速指导 presents a coherent five-step selection sequence with a clear ordering (problem classification → constraint check → complexity threshold → extension preference → check existing project examples), and 不适用场景 provides explicit disqualification criteria. No validation checkpoints exist, but this is an advisory selection skill, not a destructive or batch operation, so the cap does not apply. Not 5 because there is no explicit verification step (e.g., confirming the chosen pattern matches an existing project example) before recommending.

Progressive Disclosure

References are clearly labeled and one level deep in 延伸阅读, but none of the three referenced files ('reference/1-core-patterns.md', 'reference/2-selection-guidelines.md', 'reference/3-project-examples.md') exist — the bundle contains only SKILL.md, with no reference/ or references/ directory. The body is otherwise well-organized and short, but its core progressive-disclosure structure is broken: navigation leads nowhere. Not 3 because broken references are a worse navigation failure than unclearly-signaled ones; not 1 because the body itself has clear sections and no nested-reference chains.

Specificity

Names the domain (BK-CI design-pattern selection) and six concrete pattern names (工厂、策略、责任链、观察者、模板方法、装饰器), but the only stated action is '判断何时使用' (judge when to use) — a single action rather than several specific capabilities. Not score 4 because it does not list several distinct concrete actions; not score 2 because the domain and triggers are far more concrete than 'Processes PDF files'.

Completeness

Both parts are present: a 'what' (decide when to use which design pattern in BK-CI) and an explicit 'when' ('当用户要重构可扩展结构、统一模式选型、理解项目里的模式落点时优先使用'). The 'when' is explicit with concrete trigger phrases, but the 'what' is thin and somewhat circular ('use when judging when to use...'), so it does not clearly reach the anchor-5 bar of a fully concrete what + when.

Trigger Term Quality

Contains natural terms users would actually say: 重构可扩展结构, 统一模式选型, 理解模式落点, plus all six pattern names. Good coverage but missing common variations such as 代码坏味道, 解耦, or English terms users might mix in. Not 5 because synonym coverage is incomplete; not 3 because far more than 'some relevant keywords' are present.

Distinctiveness Conflict Risk

Scoped tightly to BK-CI and to pattern selection, with distinct triggers — clearly a niche. Minor overlap risk with the sibling skills it names in its own body ('utility-components', 'common-technical-practices') and with general refactoring guidance. Not 5 because adjacent project-practice skills could compete for refactoring-related queries.