agf-writing-qa-report

$npx mdskill add pcliangx/AppGenesisForge/agf-writing-qa-report

Publishes E2E or UAT QA reports with evidence and verdict criteria.

  • Generates report skeleton for E2E or UAT sign-off.
  • References UAT case documents as evidence source of truth.
  • Appends re-run sections to existing report files.
  • Outputs markdown report to docs/qa/ with stage and date.

SKILL.md

.github/skills/agf-writing-qa-reportView on GitHub ↗
---
name: agf-writing-qa-report
description: Use when qa-engineer (or miniapp-qa-engineer) is about to publish an E2E or UAT report. Provides the report skeleton, evidence-quality bar, verdict criteria, and hand-off rules. SIT is now dev-owned and lives in progress/<role>.md (see agf-running-sit-tests skill) — this skill does NOT cover SIT reports.
---

# Writing a QA Report (E2E / UAT)

Use this skill when:

- E2E execution is complete and a report needs publishing
- UAT business sign-off needs to be captured
- A bug-fix E2E re-verification needs recording (always **appended** as `## Re-run [N] — [date]` to the existing `[feature]-e2e-[YYYY-MM-DD].md`; never a new file — see "File path & naming" below)

**Pair with**:
- `agf-running-sit-tests` skill — SIT 由 dev 自跑,本 skill 不覆盖 SIT 报告(SIT 证据已在 `progress/<role>.md` 的 `**SIT 证据**` 段,归档随 `docs/qa/<feature>-process-log.md` 走)
- **UAT 用例文档** `docs/qa/[feature]-uat-cases-[date].md`(模板 `docs/qa/uat-cases-_TEMPLATE.md`,gate SSOT 见 `testing.md`「UAT 用例文档」节)— UAT 执行前生成 + 用户审核 `status: Approved`;执行证据回填用例文档(证据 SSOT),**UAT 报告引用用例 ID,不重复粘贴证据**
- This skill — covers the **E2E / UAT artifact** (report file) format

## File path & naming

`docs/qa/[feature-kebab-case]-[stage]-[YYYY-MM-DD].md` — Stage ∈ `{e2e, uat}`. Examples:
- `docs/qa/oauth-login-e2e-2026-05-13.md`
- `docs/qa/oauth-login-uat-2026-05-15.md`

**One report per stage per feature.** Re-runs after defect fix → append a new `## Re-run [N] — [date]` section to the same file, do not create a new file.

## Required sections (in order)

```markdown
---
# frontmatter 是 verdict 数据的唯一 SSOT(agf-verdict.py 解析;validate-verdict hook 校客观底线、agf-matrix.sh fan-in 都读这里)
feature: [feature-slug]
date: YYYY-MM-DD
tester: qa-engineer
stage: E2E                     # E2E | UAT
report_verdict: Promote        # Promote | Conditional promote | Block
critical_defect_count: 0       # 客观底线事实(>0 时 report_verdict 必须为 Block)
p0_pass2_total: 0              # P0 用例数(需 pass²)
p0_pass2_ok: 0                # 连续 2 次都过的 P0 数
uat_signoff_verdict: N/A       # approve | request changes | N/A(仅 UAT 阶段有值;不推导)
---

# QA Report — [Feature] — [E2E|UAT]

- **Date**: YYYY-MM-DD
- **Stage**: E2E / UAT
- **Tester**: qa-engineer ([model name]) / 业务方姓名(UAT)
- **Branch**: [branch + commit hash]
- **Environment**: local docker-compose / staging / pre-prod
- **需求入口**: docs/changes/[change]/tasks.md(AC↔scenario 映射,ADR-012;旧 feature fallback docs/prd/[feature]-[date].md)
- **Code review (含 SIT Audit)**: docs/reviews/[feature]-[date].md

## Summary

- Total AC: N
- Passed: M
- Failed: K
- Blocked: J
- 界面渲染核查(仅 UAT 且含界面 feature): N/N 界面真渲染 + 截图 + 读图四查通过(矩阵 SSOT 在用例文档)
- **Verdict**: ✅ Promote to next stage / ❌ Block / ⚠️ Conditional promote

## Pre-conditions Checked

- [ ] 单元测试 + lint + typecheck 全绿
- [ ] code-reviewer 报告已存在且 verdict ≠ Block(含 SIT Audit = ✅ / ⚠️)
- [ ] AC 来源可访问(`docs/changes/<change>/tasks.md`;旧 feature fallback PRD)
- [ ] 环境就绪(DB 起来 / 迁移已 apply / 服务已启动)
- [ ] **(仅 UAT)用例文档已审核**:`docs/qa/[feature]-uat-cases-[date].md` 存在且 frontmatter `status: Approved`(MAJOR / MINOR 强制;PATCH 级 hotfix 由 PL 豁免时在报告注明理由)

任何一条没勾 → 不该开始测;先 SendMessage product-lead 解决先决条件。

## AC Results

### AC-1 (P0): [verbatim AC text from tasks.md(fallback PRD)]

- **Priority**: P0 / P1 / P2(来自 `docs/changes/<change>/tasks.md` 的 AC 优先级标注;旧 feature 来自 PRD §4 Priority;P0 必须跑 2 次,P1/P2 跑 1 次即可)
- **Setup**: [起始状态]
- **Action**: [触发步骤]
- **Expected**: [复制 AC 原文]
- **Actual (run 1)**:
  ```
  HTTP/1.1 200 OK
  Content-Type: application/json
  {"id": 42, "status": "created"}
  ```
- **Actual (run 2)** [P0 必填;P1/P2 可空]:
  ```
  HTTP/1.1 200 OK
  ...
  ```
- **Reliability**: `pass^1 = 1/1` 或 `pass^2 = 2/2`(P0)— 两次不一致 = `⚠️ Flaky`,按 fail 处理
- **Verdict**: ✅ Pass / ❌ Fail / ⚠️ Blocked / ⚠️ Flaky

(每个 AC 都要单独一节,**禁止合并写 'all passed'**。**为什么 P0 要跑 2 次**:业界实证(τ-bench)pass@1 高 ≠ pass^k 高,单次过的 P0 case 偶发问题会逃逸到生产。)

## Defects Found

| ID | Severity | Title | Repro steps | Suspected file |
|---|---|---|---|---|
| DEF-1 | High | ... | 1. ... 2. ... | backend/app/foo |

Severity 标准:
- **Critical**: 阻断核心流程,无 workaround
- **High**: 阻断核心流程但有 workaround;或非核心流程的数据/安全问题
- **Medium**: 边缘场景失败,不阻断核心流程
- **Low**: 体验/文案/兼容性

## Cross-stage Notes

- E2E → UAT: 给业务方的操作手册 / 数据准备说明 / 已知 P2 defect 列表

## Cost (this QA session)

- Tokens consumed: [from `/usage`]
- Estimated cost: [CNY]
- 同 feature 累计(E2E + UAT 总和):[CNY]

## Hand-off

✅ Promote → SendMessage product-lead 进下一阶段
❌ Block → SendMessage product-lead 列 critical defect,重新派回 dev
⚠️ Conditional → 列 P2 defect 单独建 issue,allow 进下一阶段
```

## Verdict 决策树(不能凭感觉)

```
任一 P0 AC = Fail              → ❌ Block
所有 P0 + P1 AC = Pass         → ✅ Promote
P0 全 Pass,P1 部分 Fail        → ⚠️ Conditional(P1 失败必须建跟踪 issue)
有 P0 = Blocked(环境问题)     → ⚠️ Block + 升级 product-lead
```

### 客观底线硬校验(frontmatter,退出时 hook 拦)

`report_verdict` 的定性选择(Promote vs Conditional)**不**套公式,但两条**客观底线**由 `validate-verdict.sh`(委托 `agf-verdict.py`,ADR-010)退出时重算、违反即 exit 2 打回——填 frontmatter 前自查:

- `critical_defect_count > 0` → `report_verdict` **必须为 Block**(有 critical defect 却 Promote/Conditional = 不可能组合)。
- `p0_pass2_ok < p0_pass2_total`(P0 未全部 pass²)→ `report_verdict` **不得为 Promote**(P0 没连过 2 次不准晋级)。

> 这两条只拦"客观不可能组合",不强定 Promote/Conditional 档;极保守 fail-open(缺字段 / 坏 yaml 一律放行 + WARN)。`uat_signoff_verdict` 仅结构化记录、不推导。

## Evidence 质量条

每条 Pass 的 Actual 段**必须**有可验证产物之一:

- HTTP 调用:`curl -i` 完整响应头 + body(敏感字段可遮)
- DB 状态:`SELECT` 前后对比
- UI:截图(命名规则:`evidence/AC-N-[step].png`)
- 日志:相关行(带时间戳)
- 文件落盘:`ls -la` + 内容 head

**禁止**只写"Passed, looks correct"——这种 Pass 不可信,等同没测。

**UAT 含用户可见界面的用例额外强制**(SSOT:`testing.md`「UAT 界面渲染核查」节):

- 截图**必选**且必须来自 chrome-devtools 真渲染(小程序 / Apple 轨用对应模拟器;命名 `evidence/UAT-[case]-[界面slug].png`),**落盘后必须用 Read 读回做视觉分析**(对照 design spec + `index.html` 原型),在用例文档「界面渲染核查矩阵」回填四查结论(导航 / 裁切 / 控件可点 / 视觉达标)
- curl / `SELECT` 输出只能作**补充**——**纯 API / DB 断言不构成 Pass**;矩阵任一行缺截图或缺读图结论("已截图" ≠ "已核查")= 该界面未测,本报告不得发布

## 反模式

("合并写 all passed" / "无 evidence 的 Pass" / "Defect 不写 Repro" 已由下文「完成前的验证」checklist 反向守门,此处不重列。)

- ❌ 跑 E2E 时同时改代码(移动靶)— 必须 freeze 分支再跑
- ❌ 用生产 API key 跑 E2E — 必须用专用测试 key + 每日花费上限
- ❌ 用本 skill 写 SIT 报告 — SIT 已 dev 自跑,证据落 `progress/<role>.md`,不再有独立 SIT 报告
- ❌ UAT 用 API 断言代替界面渲染 — 接口 200 ≠ 界面可用(导航缺失 / 裁切 / 控件点不动全漏检);含界面用例必须真渲染 + 截图 + 读图四查,"时间紧 / E2E 已截过图 / 后端数据对了"都不是豁免理由
- ❌ 截图存档但不读图 — 截图必须用 Read 读回、以视觉能力对照 design spec 给"视觉达标"结论;**UIUX 是用户对产品最直接的感受**,"功能对了但界面糙"不是可交付状态,"已截图"不等于"已核查"

## 完成前的验证

- [ ] 每条 AC 都有 Setup / Action / Expected / Actual / Verdict 五段?
- [ ] 每个 Pass 都有可验证 evidence?
- [ ] Defects 表每行都有 Repro steps + Suspected file?
- [ ] Cost 一节填了实际数字(不是 TBD)?
- [ ] Verdict 由决策树推出(不是凭感觉)?
- [ ] **交互控件全覆盖**(含前端的 feature):页面每个可交互控件都点击/输入过 + 断言了可观测后果(DOM/网络/路由/状态),非"截图看着有按钮"(见 `testing.md` 前后端对接强制覆盖项 ③)?
- [ ] **(仅 UAT)界面渲染核查矩阵**:每个用户可见界面已真渲染 + 截图 + 读图四查(导航 / 裁切 / 控件可点 / 视觉达标),每张截图都被 Read 读回分析过,无"待执行"残留、无以纯 API 断言代替(见 `testing.md`「UAT 界面渲染核查」节)?
- [ ] Hand-off SendMessage 已发出?

任一不行 → 不要 publish,回去补。

## Hand-off 触发

报告落盘后立即(**不等用户问**):

1. SendMessage product-lead,附 verdict + report path + 1 句话总结
2. 如 verdict = Block:列 top-3 critical defect 在消息正文
3. 如 verdict = ⚠️ Conditional:把 P2 defect 通过 `TaskCreate` 单独开 follow-up task(由 product-lead 派发)

More from pcliangx/AppGenesisForge

SkillDescription
agf-code-mapDeeply Understand (codemap) — 持久化代码图谱 + 变更影响分析 + 理解地图。接手遗留项目 Day-1 / PR 影响分析 / 解释陌生代码 / PRD·ADR 前现状理解时用。编排 tools/codemap/ 的 codemap CLI
agf-deploying-uatUse when deploy-engineer is about to deploy the merged-to-main code to the isolated local UAT stack (after code review + SIT Audit pass and merge, before qa-engineer runs E2E/UAT). Provides the applicability gate, pre-flight checks, isolated compose bring-up (independent project name + port offset +900), in-container migration, real-output smoke test, hand-off, and the deploy-report skeleton. Pairs with deployment.md "UAT 环境部署" contract and slash /agf-deploy-uat.
agf-design-disciplineUse when uiux-designer is about to produce a design spec (spec.md) or static HTML prototype, or frontend-dev is about to build UI from a design. Provides the anti-AI-slop design discipline layer — Brief Inference (Design Read), three aesthetic dials tuned for product UI, AI Tells blacklist with overrides, mechanically-checkable Pre-Flight. Sits above the token layer (DESIGN.md) and mechanical review (code-reviewer); does not redeclare tokens or guide non-shadcn design systems. Inspired by taste-skill, cropped for AGF product UI per ADR-013.
agf-releasing-appleUse when apple-release-engineer is about to build the signed distributable (TestFlight build / notarized DMG / internal package) from merged-to-main code (after apple code review + SIT Audit pass and merge, before apple-qa-engineer runs E2E/UAT). Provides the applicability gate, pre-flight checks, lane execution per channel, notarization, real-output smoke test, hand-off, and the release-report skeleton. Pairs with deployment.md §7 "Apple 发布" contract and slash /agf-apple-release.
agf-running-apple-sitUse when apple-dev has finished feature code + Unit tests (Swift Testing) and is about to enter code-review. Provides the Apple SIT scope (xcodebuild test + simulator per declared target), the AC-driven integration walk, APIProtocol-mock discipline, and evidence sink (progress/apple-dev.md). SIT is dev-owned; apple-code-reviewer audits the evidence.
agf-running-release-retroUse when product-lead is about to run a release retrospective after a successful MAJOR or MINOR release push (PATCH skipped). Provides applicability gate, pre-conditions, 7-step execution sequence, anti-patterns, and the verification gate before commit. Pairs with template docs/reviews/retro-_TEMPLATE.md and slash /agf-release-retro.
agf-running-sit-testsUse when an execution-layer dev (frontend-dev / backend-dev / ai-agent-dev / ml-engineer / miniapp-dev) has finished feature code + Unit tests and is about to enter code-review. Provides the SIT scope, environment, AC-driven integration walk, and evidence sink (progress/<role>.md). SIT is now a dev-owned step, not a separate QA stage.
agf-skill-stocktakeUse when PL / tech-lead audits the content quality of AGF's own skills (`.claude/skills/agf-*/SKILL.md`). Provides the 5-dimension checklist (Actionability / Scope fit / Uniqueness vs CLAUDE.md+standards / Currency of pointers / Gate-binding validity), 5-verdict rubric (Keep/Improve/Update/Retire/Merge), reason-quality bar (反敷衍), and output skeleton. Triggered each MINOR release retro or manually.
agf-wiring-apple-llmUse when wiring LLM features into the Apple client (macOS / iOS) — streaming chat via the FastAPI multi-LLM gateway, or on-device inference (Apple Foundation Models / Core ML). Provides the route decision (gateway vs on-device), streaming transport pattern, env/config contract, offline & cost guardrails, and minimum verifications before declaring the integration done.
agf-wiring-multi-llm-sdkUse when wiring up or switching between China-domestic LLM providers (DeepSeek, Doubao/Volc Ark, Qwen/DashScope, MiniMax). Provides OpenAI-compatible adapter pattern, env-var contracts, fallback strategy, cost guardrails, and minimum verifications before declaring integration done.