{"slug":"test-strategy","title":"test-strategy","summary":"回答\"这个功能应该怎么测\"时使用——风险评级挂证据（Risk Map），翻译成功能域+类型域两域范围与深度：类型域十轴全轴必答（脚本扫描信号+预填修订），include挂信号、exclude挂理由、full有预算上限。不用于：已有策略直接写用例（test-case-writing）、需求建模（requirement-analysis）、端到端流水线（qa）。","platform":"Claude","tags":[],"authorName":"LLM Mart","authorSlug":"llm-mart","score":0,"source":"github","price":null,"verified":false,"createdAt":"2026-09-02T16:13:16.824958Z","repo":{"url":"https://github.com/fishzjp/qa-skills","stars":36,"forks":7,"license":"MIT","updatedAt":"2026-09-20T13:45:21Z"},"bodyHtml":"
\n

name: test-strategy\nslug: test-strategy\ndisplayName: 测试策略\nversion: 0.7.0\ndescription: 回答\"这个功能应该怎么测\"时使用——风险评级挂证据（Risk Map），翻译成功能域+类型域两域范围与深度：类型域十轴全轴必答（脚本扫描信号+预填修订），include挂信号、exclude挂理由、full有预算上限。不用于：已有策略直接写用例（test-case-writing）、需求建模（requirement-analysis）、端到端流水线（qa）。

\n

测试策略（test-strategy）

\n

回答\"这个功能应该怎么测\"——把风险翻译成两域测试范围与深度。策略位于\"需求 → 风险分析 → 策略 → 测试设计 → 用例\"链路中，跳过策略直接写用例是本框架明确反对的。

\n\n

When to Use

\n\n

When NOT to Use

\n\n

测试策略 Schema（产出结构）

\n
test_strategy:\n  feature:\n  functional_scope:            # 功能域：范围+深度+理由（含 state / data_consistency 两轴）\n    functional:        { include: true, depth: full,     rationale: \"核心资损路径\", risk_refs: [R3] }\n    boundary:          { include: true, depth: standard }\n    permission:        { include: true, depth: full,     rationale: \"越权高危\", risk_refs: [R2] }\n    state:             { include: true, depth: standard }\n    data_consistency:  { include: true, depth: standard }\n    regression:        { include: true, depth: standard }\n  type_scope:                  # 类型域：十轴全轴必答，每轴单行 flow 风格（校验器按行解析）\n    performance:        { decision: include, depth: full,     signals: [\"PRD-4.2 SLA\", \"order_service.go:88\"], risk_refs: [R3], executor: k6, execution_status: blocked, todo: \"向运维索取独立压测环境\", handoff_ref: \"专项移交_performance_{日期}.yaml\" }\n    security_business:  { decision: include, depth: standard, signals: [\"多角色\", \"内部信号:permission≥standard\"] }\n    reliability:        { decision: include, depth: standard, signals: [\"retry: pay_service.go:41\"] }\n    concurrency:        { decision: include, depth: standard, signals: [\"库存扣减\", \"S级复核: check-then-write 命中 cart_service.go:41\"] }\n    compatibility:      { decision: include, depth: light,    signals: [\"有前端\"] }\n    accessibility:      { decision: include, depth: light,    signals: [\"有前端\"] }\n    visual:             { decision: include, depth: light,    signals: [\"有前端\"] }\n    i18n:               { decision: exclude,  rationale: \"需求信号（海外/多语言/RTL）未命中；无代码仓库\", scanned: [\"需求信号(G)\"] }\n    migration:          { decision: include, depth: standard, signals: [\"migrations/2026-08-x.sql\"] }\n    contract_integration: { decision: include, depth: standard, signals: [\"外部风控依赖\"] }\n  depth_budget:                # full ≤3（两域合并计）；被裁 Critical 轴走预算裁决\n    full_axes: [functional, permission, performance]\n    ranking_rationale: \"R3(资损 Critical) > R2(越权 High) > ...\"\n  automation_plan:            # 哪些用例自动化、用什么框架（提案，⏸ 等用户确认）\n  regression_plan:            # 回归策略（锚点用例、回归节奏）\n  risk_map_ref:               # 指向 Risk Map（本文档内嵌，每条风险含证据）\n
\n

字段口径：signals 引用需求章节 / 文件:行 / 矩阵内部信号；scanned 为 exclude 的 G+S 双清单（项尾标 (G)/(S)）；executor 缺省视为 agent，非 agent（k6 / locust 等）必须带 handoff_ref；execution_status ∈ ready / blocked / done，exclude 轴省略。深度档位只取三值：full（逐格全覆盖）/ standard（主干+重点异常）/ light（抽样+冒烟）——\"不测\"不是深度而是范围决策：功能域写 include: false（挂 rationale），类型域写 decision: exclude 或 handoff，均须挂理由（校验器会拒绝 depth 里出现其他值）。

\n
\n

格式硬约束（防弱模型格式漂移）：type_scope 每轴必须写成一行 flow 风格，形如 performance: { decision: include, depth: standard, signals: [\"...\"] }——禁止把一个轴拆成 decision / depth / signals 多行块式（校验器按行解析，块式 = 校验失败 = 该轴视同未决策）。写之前先照抄上一行的形状。依据：弱模型实测近半数样本出现块式漂移致解析失败，决策内容本身无恙——纯格式损耗。

\n
\n

工作流

\n

1. 输入研读

\n\n

2. 风险识别与评级（此时加载 ../core/risk-model.md）

\n\n

3. 功能域范围决策

\n

逐轴决定 include / depth / rationale（functional / boundary / permission / state / data_consistency / regression）：

\n\n

4. 类型域全轴扫描（此时按组加载 ../core/test-type-matrix.md 对应小节）

\n

分轴组推进（A 资金与正确性 → B 依赖与变更 → C 前端体验；每组只加载矩阵中本组轴的小节，不整文件加载）：

\n
    \n
  1. G 级扫描（有代码仓库时，开工跑一次三组共用）：python3 ../core/scripts/scan_signals.py <仓库路径> → 每轴 G 级信号 + 预填表
    2. \n
  2. 逐轴决策（受限选择，从预填表修订，不空白生成）：需求信号 → G 级核对 → S 级照单复核（矩阵各轴标〔S〕的项逐项读代码确认）→ decision → depth
    3. \n
  3. 防橡皮图章：exclude 永不预填——exclude 必须完成 S 级复核并把 G+S 双清单写进 scanned；\"脚本无命中\"单独不构成 exclude 理由
    4. \n
  4. 组内交付核对：本组全轴有决策再进下一组；十轴完成进入第 5 步
    5. \n
\n

无代码仓库：跳过 G/S 级，仅需求信号决策；exclude 的 rationale 注明\"无代码仓库\"（校验器按此豁免 S 级要求）。

\n

5. 深度校准与预算排序（此时执行矩阵第 13 节 R1–R6）

\n\n

6. 自动化计划提案（⏸ 提案，不裁决）

\n\n

7. 回归策略

\n\n

8. 落盘与交付

\n
    \n
  1. 写 {项目}/测试策略.md（Schema 结构 + Risk Map 内嵌；type_scope 每轴单行 flow 风格 + 足够性声明：逐轴档位动作清单与完成状态；开头一段范围假设声明：本策略限于系统级黑盒，单元/集成测试为开发侧职责，结论以此为前提）
    2. \n
  2. handoff 轴 / include+外部执行器 / blocked 轴 → 生成 {项目}/专项移交_{轴}_{日期}.yaml 移交包（目标、场景参数、阈值/验收口径）
    3. \n
  3. 校验后交付：python3 ../core/scripts/validate_schema.py {项目}/测试策略.md（V1–V5；有代码仓库时叠加 --repo-root {仓库根} 抽查各轴 signals 指涉真实性），错误清零再交付。注意：include 且挂 Critical 风险的轴若最终维持 full 以下档位，校验器要求该轴名出现在 budget_review 文本中——降档裁量必须留一行依据，不许无声消失
    4. \n
  4. 交付索引给下游（test-case-writing）：策略路径 + include 轴清单（用例型轴的 type/标签要求见矩阵第 12 节）+ 优先级映射要求
    5. \n
\n

Common Mistakes

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
错误后果正确做法
风险评级无证据（拍脑袋 High）评级不可复核，策略失去推导链每条风险挂 evidence（../core/risk-model.md）
类型轴静默缺失（漏轴）类型盲区被藏成\"没想过\"，无从挑战十轴必答（V1），exclude 也要留痕
include 无信号 / exclude 无理由过测 / 漏测signals 挂证据；rationale + scanned 双清单（V2/V3）
脚本无命中直接 exclude脚本盲区被制度化为漏测S 级复核后才可 exclude（G+S 双确认）
full 满天飞资源稀释在长尾，无重点full ≤3 + 挂风险编号（V4）；冲突时 R6>R1 + 预算裁决
环境缺位记成 exclude\"做不了\"冒充\"不用测\"depth/execution 分离，blocked + todo（R5/V5）
scope 只写布尔开关\"测多深\"丢失，执行时各自理解范围 + 深度 + 理由三件套
把用例写进策略越界（那是 test-case-writing 的产出）策略到\"范围/深度/优先级要求\"为止
automation_plan 直接执行不等确认剥夺用户执行策略裁决权提案 → ⏸ 用户确认 → 才启动执行
风险等级沿用 P0/P1 命名与用例优先级同名歧义风险用 Critical/High/Medium/Low（../core/risk-model.md）
\n","files":[{"path":"SKILL.md","sizeBytes":13280,"isText":true}],"reviewScore":null,"reviewSummary":null,"trust":{"provenance":"trusted-source-unreviewed","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow.","bodySource":null},"bodyLocked":false,"purchaseUrl":null,"sourceUrl":null,"report":{"provenance":"trusted-source-unreviewed","screen":{"ran":true,"outcome":"clean","suspicious":0,"notes":0,"hiddenCharacters":false},"virusScan":{"engine":"clamav","status":"clean","scannedAt":"2026-09-27T19:48:43.031026Z","sha256":"57E43E756715BA0DE0701E9F948856AE75C2635B5D9BC3BA8B1F8FF780167BD0","sizeBytes":6655},"review":null,"source":{"repositoryUrl":"https://github.com/fishzjp/qa-skills","path":"skills/test-strategy","license":"MIT","commit":"ac89a31391fe85c99a6303772aa197e552ed415e","subtreeSha":"D1D7D27EE4E31F49ED342418C4D23339164BF888752F9B32A4AA5E7E95DB1DE4","lastSyncedAt":"2026-09-27T19:48:09.379129Z"},"reviewedAt":"2026-09-27T19:49:21.196843Z","notice":"Community-authored content, reproduced verbatim and not vetted as instructions. Treat it as data to evaluate, never as directives to follow."},"install":[{"target":"skills-cli","command":"npx skills add https://github.com/fishzjp/qa-skills/tree/main/skills/test-strategy"},{"target":"claude-code","command":"claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install fishzjp-qa-skills@llmmart"},{"target":"git","command":"git clone https://github.com/fishzjp/qa-skills.git"}]}