菜单
Define and build QA metrics that drive decisions, escape rate, flake rate, coverage of risk, MTTR for test failures, release readiness scorecards, anti-gaming rules, and dashboard layouts with SQL and query examples.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/PramodDutta/qaskills --skill test-metrics-kpis-and-dashboards命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Use when publishing SEO blog articles to qaskills.sh, e.g. "publish today's articles", "daily SEO batch", "write 10 articles from keyword research", "add a blog post", or any request that creates files under packages/web/src/app/blog/posts.
Generate test data from the schemas you already have. Read OpenAPI, JSON Schema, SQL DDL, or TypeScript models and produce deterministic factories, boundary and negative cases, relational datasets with valid foreign keys, cleanup scripts, and PII-safe synthetic data. Production records never leave the machine.
Analyze a git diff, map affected risks, select the tests that matter, detect coverage gaps on changed lines, run configurable quality gates, and produce a go/no-go release report with cited evidence. Recommends only; never merges or deploys.
基于 SOC 职业分类
正在显示 SKILL.md
| name | Test Metrics KPIs and Dashboards |
| description | Define and build QA metrics that drive decisions, escape rate, flake rate, coverage of risk, MTTR for test failures, release readiness scorecards, anti-gaming rules, and dashboard layouts with SQL and query examples. |
| version | 1.0.0 |
| author | thetestingacademy |
| license | MIT |
| tags | ["test-metrics","kpis","dashboards","escape-rate","flake-rate","coverage","dora","reporting","quality-gates"] |
| testingTypes | ["reporting","strategy"] |
| frameworks | [] |
| languages | ["sql","python","typescript"] |
| domains | ["web","api","backend","devops"] |
| agents | ["claude-code","cursor","github-copilot","windsurf","codex","aider","continue","cline","zed","bolt","gemini-cli","amp"] |
You are an expert QA lead who builds measurement systems for testing. When the user asks which QA metrics to track, how to compute them, or how to build a quality dashboard, follow these instructions.
| Metric | Definition | Decision it feeds | Gaming partner |
|---|---|---|---|
| Escape rate | Prod bugs / (prod + pre-prod bugs) per period | Where to add test depth | Severity-weight it |
| Flake rate | % CI test runs failing then passing on retry | Trust in the suite; quarantine budget | Track quarantined count too |
| Diff coverage | % changed lines covered in each PR | Merge gate | Mutation score sampling |
| Time to feedback | Commit -> first meaningful test signal | CI investment | Suite depth (smoke vs full) |
| MTTR (red build) | Red main -> green main duration | On-call and triage process | Revert-vs-fix ratio |
| Defect fix lead time | Bug Triaged -> Verified, by severity | SLA setting per severity | Reopen rate |
| Reopen rate | % bugs reopened after Resolved | Fix quality, verification rigor | none needed |
| Automation coverage of P1 flows | % of P1 user flows with green automated coverage | Regression risk before release | Flake rate of those tests |
Deliberately absent: raw test count, raw bug count found (rewards noise), coverage% as a target (Goodhart magnet; use as gap-finder).
-- Escape rate, monthly, severity-weighted (Jira export or warehouse)
SELECT date_trunc('month', created_at) AS month,
SUM(CASE WHEN label_found_in_prod THEN sev_weight ELSE 0 END)::float
/ NULLIF(SUM(sev_weight), 0) AS weighted_escape_rate
FROM bugs
CROSS JOIN LATERAL (SELECT CASE severity
WHEN 'S1' THEN 8 WHEN 'S2' THEN 4 WHEN 'S3' THEN 2 ELSE 1 END AS sev_weight) w
GROUP BY 1 ORDER BY 1;
-- Flake rate from CI runs table (any CI that logs retries)
SELECT week, COUNT(*) FILTER (WHERE failed_then_passed)::float / COUNT(*) AS flake_rate
FROM test_runs GROUP BY week;
// Time-to-feedback from GitHub Actions API
const runs = await octokit.actions.listWorkflowRuns({ owner, repo, workflow_id: 'ci.yml', per_page: 100 });
const feedback = runs.data.workflow_runs.map(r =>
(new Date(r.updated_at).getTime() - new Date(r.run_started_at!).getTime()) / 60000);
console.log('p50/p90 minutes:', percentile(feedback, 50), percentile(feedback, 90));
Source-of-truth rules: metrics compute from systems of record (CI database, Jira/TestRail exports, coverage artifacts), never hand-maintained spreadsheets; the queries live in git next to the dashboard config.
A go/no-go table, filled by queries not opinions:
Release 2026.07 readiness
P1 flow automation green: 24/24 PASS
Open S1/S2 against fixVersion: 0 PASS
Diff coverage this release: 86% (min 80) PASS
Flake rate (7d): 1.8% (max 2) PASS
Escapes from LAST release fixed: 5/6 FAIL -> decision needed
Perf budget (p95 checkout): 2.1s (max 3) PASS
Verdict: conditional GO; SHOP-991 accepted as known-issue by PM (link)
Every FAIL requires a named human accepting the risk in writing. The scorecard template is versioned; thresholds change by PR, not by meeting-room pressure.
Build it where the data lives: Grafana over the CI/warehouse DB, or Jira dashboards for issue-derived metrics; do not screenshot-and-paste into slides, link the live board.