AI Testing Skills: The 27 I Use Every Day, Packed!

Someone once asked me: what exactly is the Agent Skill I keep talking about, and which ones do you usually use?

There are many kinds of Skills I use in daily work: office efficiency, creative work, automated workflows, and so on. Added up, there are quite a few. I'll share those another time. Today I'll focus on the ones strongly related to AI testing.

Actually, two months ago I already shared an article about AI Testing Skill Encyclopedia, but many readers still ask in the comments, and I've added several new Skills since then, so here is another share.

For AI testing, I use a main set of 27 Skills, and all 27 were built by me. They cover the complete software testing process: requirement decomposition and test case design, script generation, automated testing, end-to-end business scenario testing, all the way to test reports and CI pipelines. Each one is built around what enterprise teams actually use, not for gathering dust in a favorites folder. You can get the Skills at the end.

First, here is the full panorama of AI testing Skills.

skills/
├── req-to-user-story/           # Requirements → User stories
├── review-user-stories/         # User story review
├── generator-testcase-xmind/    # Test case XMind generator
├── generator-testcase-excel/    # Test case Excel generator
├── review-testcase/             # Test case quality review
├── safe-testcase/               # Test scenario omission completion
├── api-schema-parser/           # API definition parser
├── api-testdata-generator/      # Automated test data construction
├── api-testscript-generator/    # API automated test script generation
├── api-test-optimizer/          # API automated script quality check and optimization
├── api-test-tagger/             # API test script intelligent tagging
├── api-test-executor/           # API test intelligent execution and scheduling engine
├── api-failure-diagnoser/       # Test failure intelligent diagnosis and auto-repair
├── api-testdata-cleaner/        # API test data cleanup tool
├── api-report-generator/        # API test intelligent report generation expert
├── api-pipeline-scheduler/      # API automated full-chain pipeline scheduler
├── api-scenario-modeler/        # Multi-API business scenario modeling
├── api-scenario-runner/         # Multi-API business scenario chain execution engine
├── ui-page-parser/              # UI page parser
├── ui-testscript-generator/     # UI test script generation
├── ui-testscript-enhancer/      # UI test script enhancement
├── ui-visual-assert/            # UI visual assertion
├── ui-test-tagger/              # UI test script intelligent tagging
├── ui-test-executor/            # UI test intelligent execution and scheduling engine
├── ui-failure-diagnoser/        # UI test failure diagnosis and auto-repair
├── ui-report-generator/         # UI test report generation expert
├── ui-pipeline-scheduler/       # UI automated full-chain pipeline scheduler
└── ...                          # More skills continuously added

The 27 Skills are grouped by process into three groups: 6 for requirements and test case design, 12 for API automation, and 9 for UI automation.

Each group can be chained together; every link in the chain can also be used independently and can automatically connect with upstream and downstream links.

1. Requirements and Test Case Design, 6 Skills

1. req-to-user-story

Requirements to user story. It automatically splits colloquial, messy raw requirements into structured user stories, identifying roles, main flow, alternative flow, exception scenarios, and pre/post conditions. Just paste in the requirements document.

2. review-user-stories

User story review. It gives the split user stories a systematic check, specifically looking for missing scenarios and missing dimensions. If you're not confident after splitting, run this over them.

3. generator-testcase-xmind

Test point generation. Based on user stories, it splits test points and outputs an XMind mind map covering five dimensions: functionality, boundary values, exceptions, business rules, and non-functional requirements, with priorities marked P0 to P2.

4. generator-testcase-excel

Test case generation. It turns requirement user stories or XMind test points into formatted Excel test cases, complete with number, preconditions, steps, test data, expected results, and related requirements. They can go directly into the case library.

5. review-testcase

Test case quality review. It grades AI-generated cases (usable, needs modification, invalid/error), scores them on five dimensions, calculates seven quantitative metrics, and gives per-case deduction reasons and improvement suggestions. This is the final quality gate for AI-generated cases.

6. safe-testcase

Scenario omission completion. It specifically digs out the scenarios humans most easily miss: historical bug reviews, non-functional, security, compatibility, and hidden business rules. Put in existing cases, and it fills the gaps and marks them.

2. API Automation, 12 Skills

7. api-schema-parser

API definition parsing. Swagger, Postman, HAR capture, YApi export, or plain-text description, no matter the format or source, it uniformly parses into a standard API definition file. All later links share this one data source.

8. api-testdata-generator

Test data construction. Pick any of three modes: API definition, field rules, or natural language. It automatically generates test data across four dimensions: positive, boundary, exception, and security.

9. api-testscript-generator

API script generation. Based on the API definition, it batch-generates a layered script project using Python + Requests + Pytest + Allure, data-driven, and runnable as soon as it's generated.

10. api-test-optimizer

Script quality optimization. It performs four kinds of validation on generated scripts: syntax, conventions, robustness, and logic, then supplements missing scenarios across ten dimensions. It outputs a validation report plus an optimized executable script.

11. api-test-tagger

Script tagging. A five-dimensional tag system (priority, module, scenario, execution strategy, environment) batch-tags scripts and automatically detects tag conflicts. From then on, select cases by tag instead of flipping through files.

12. api-test-executor

Execution scheduling engine. Run tests by speaking plain language, for example, "run smoke tests" or "run the P0 login module in the test environment." Natural language defines the scope and outputs structured execution results.

13. api-failure-diagnoser

Failure diagnosis and repair. It classifies failed cases into four categories: environment, data, script, and product defect. Script problems directly generate repair patches, which are automatically verified after modification.

14. api-testdata-cleaner

Test data cleanup. It cleans the database, Redis cache, and local temporary files in one pass. It supports scheduled CI invocation and enforces interception in production, so accidental deletion is blocked.

15. api-report-generator

Report generation. It aggregates execution results, diagnosis conclusions, and historical data, and produces a visual HTML report with risk grading, high-frequency failing API identification, and optimization suggestions. It also links with Allure for dual reports.

16. api-pipeline-scheduler

Full-process orchestration. It is the conductor that strings together execution, cleanup, and report generation in one sentence. If a single link fails, you can configure whether to continue.

The full-chain pipeline finished executing, and all links succeeded.

17. api-scenario-modeler

Scenario modeling. It converts natural-language business chains such as "register → login → add to cart → place order → pay" into standardized scenario_xxx.yaml orchestration files. All APIs are matched against api_definitions.json and never fabricated; dynamic values such as tokens and order numbers are always variable-injected, not hard-coded; and uncertain points where AI is not confident are honestly listed for human confirmation.

18. api-scenario-runner

Chain execution engine. It actually runs the orchestration file, sends requests in order, automatically extracts and injects variables, uniformly injects auth headers, and adds layered assertions plus final database assertions. On failure it stops and preserves the complete scene, handing off to api-failure-diagnoser. It can also compile to a pytest chain script and integrate into the execution scheduling system, so single APIs and business scenarios run in the same system.

After execution, it automatically generates an end-to-end full-chain test report:

3. UI Automation, 9 Skills

19. ui-page-parser

Page parsing. Given a URL or a piece of test case description, it outputs a standardized pages.yaml page object definition. It uses three-level degradation: Playwright dynamic capture, CDP connection, then inference fallback. It is the data starting point for the whole UI chain.

20. ui-testscript-generator

UI script generation. pages.yaml plus business cases generates standard Playwright + POM + Pytest scripts. It only generates what the cases involve and writes no wasted code.

21. ui-testscript-enhancer

Script enhancement. It specifically fixes flaky scripts: smart waits, captcha recognition, popup handling, iframe switching, exception retry, and failure screenshot/video capture are all added at once. It is a lifesaver for flaky cases.

22. ui-visual-assert

Visual assertion. It adds visual regression and cross-browser capability to scripts: pixel-level screenshot comparison, dynamic region masking, and three-engine difference tolerance. Questions like "did the page layout break after the redesign?" and "is the mobile layout correct?" are left to it.

23. ui-test-tagger

UI script tagging. A six-dimensional tag system batch-tags scripts, parses script semantics, and automatically recommends tags. Selecting cases by module or browser becomes one sentence.

24. ui-test-executor

UI execution scheduling. It handles tag filtering, automatic browser environment detection, and full execution monitoring. Screenshots, recordings, Trace, and network logs are automatically collected, so no failure scene is missed.

25. ui-failure-diagnoser

UI failure diagnosis and repair. It locates six failure categories and fourteen root causes. Locator drift is auto-repaired by comparing against pages.yaml to find lost elements. The red line is not touching the assertion semantics of the test case.

26. ui-report-generator

UI report generation. A single-file HTML report with status distribution, module pass rate, browser matrix, and risk grading. Failure details let you open screenshots, video, and Trace in one click, and root cause clustering gives direct conclusions.

27. ui-pipeline-scheduler

UI full-process orchestration. It automatically strings together execution, diagnosis, retry, merge, and report in one pass, with a circuit breaker fallback. "One-click run all UI automation" is what it does.

4. How These 27 Skills Are Chained Together

Each skill is an independent tool on its own; chained together, they form a complete production line.

Requirements Analysis and Test Case Design Flow

New requirements pass through the first 6 Skills: user stories, test points, test cases, review, and gap filling. The generated test cases then feed the API and UI pipelines.

Raw requirement
  │
  ▼
req-to-user-story ──→ Structured user stories
  │
  ▼
review-user-stories ──→ User story review (fill omissions)
  │
  ├──→ generator-testcase-xmind ──→ XMind test points
  │
  ├──→ generator-testcase-excel ──→ Excel test cases
  │
  ▼
safe-testcase ──→ Scenario omission completion
  │
  ▼
review-testcase ──→ Test case quality review (score + grade)

API Automated Testing Full Flow

On the API line, starting with Skill 7: API definition parsing, data generation, script generation, optimization, tagging, then the execution engine. Failures are diagnosed and repaired, data is cleaned up after running, and reports are generated. Or you can just call the pipeline scheduler in one sentence and get it all done. There is also a business scenario line: Skills 17 and 18 turn the "register → add to cart → place order → pay" business chain into an orchestration file and execute it for real. Single API tests test a point; business scenarios string together a line. Both lines share the same execution, diagnosis, and reporting system.

API docs (Swagger/Postman/HAR etc.)
  │
  ▼
api-schema-parser ──→ Standardized API data (api_definitions.json)
  │
  ├──→ api-testdata-generator ──→ Full-scenario test data
  │         │
  │         ▼
  │   api-testscript-generator ──→ API automated test script project
  │         │
  │         ▼
  │   api-test-optimizer ──→ Script quality check and optimization
  │         │
  │         ▼
  │   api-test-tagger ──→ Intelligent tag management
  │         │
  │         ▼
  │   api-test-executor ──→ Intelligent execution scheduling
  │         │
  │         ├──→ api-failure-diagnoser ──→ Failure diagnosis and auto-repair
  │         │
  │         └──→ api-pipeline-scheduler ──→ Full-chain pipeline scheduling
  │                 │
  │                 ├── api-test-executor (run tests)
  │                 ├── api-testdata-cleaner (clean data)
  │                 └── api-report-generator (generate report)
  │
  ├──→ api-scenario-modeler ──→ Business scenario orchestration file (scenario_xxx.yaml)
  │         │
  │         ▼
  │   api-scenario-runner ──→ Real chain execution (three-tier assertion + final DB check)
  │         │
  │         ├──→ run mode: execute directly and produce report, full failure scene preserved
  │         └──→ compile mode: compile pytest chain script and integrate into api-test-executor scheduling
  │
  └──→ can also directly enter generator-testcase-xmind/excel to generate API-level test cases

UI Automated Testing Full Flow

On the UI line, starting with Skill 19: page parsing, script generation, enhancement, visual assertion, tagging, then execution, diagnosis, and reporting form a chain. ui-pipeline-scheduler runs the whole thing in one click.

Page URL / test case description
  │
  ▼
ui-page-parser ──→ Standardized page object (pages.yaml)
  │
  ▼
ui-testscript-generator ──→ Playwright + POM + Pytest test scripts
  │
  ▼
ui-testscript-enhancer ──→ Script robustness enhancement (smart wait/captcha/popup/retry)
  │
  ▼
ui-visual-assert ──→ Visual regression + cross-browser + responsive testing
  │
  ▼
ui-test-tagger ──→ Six-dimensional intelligent tag management (priority/module/scenario/page/strategy/browser)
  │
  ▼
ui-test-executor ──→ Intelligent execution scheduling (tag filter/browser matrix/failure artifact collection)
  │         │
  │         ├──→ ui-failure-diagnoser ──→ Failure diagnosis and auto-repair (6 categories + 14 root causes)
  │         │
  │         └──→ ui-pipeline-scheduler ──→ Full-chain pipeline scheduling (execute→diagnose→retry→report)
  │                 │
  │                 ├── ui-test-executor (first round)
  │                 ├── ui-failure-diagnoser (diagnose and repair)
  │                 ├── ui-test-executor (retry failed cases)
  │                 ├── merge_reports (merge multi-round XML)
  │                 └── ui-report-generator (generate final report)
  │
  └──→ sub-skills can also be called independently:
        - ui-test-executor: run tests alone
        - ui-failure-diagnoser: diagnose existing failures alone
        - ui-report-generator: generate report from JUnit XML

27 Skills weld a test pipeline from one end to the other.

5. Where to Download the Skills

All the self-developed Skills above are uploaded to GitHub for anyone to take as needed.

At the same time, all Skills are now uploaded to the Skill store on testfather.cn.

The address is https://testfather.cn/skills. After registering and logging in, you can view them one by one. Members of the knowledge planet "AI 测开进化圈" get the full Skill set upon joining, and new testing efficiency Skill packs will be continuously updated.

Use Claude Code, Codex, WorkBuddy, or any other tool that supports Agent Skills. Put the Skill into the project's skills directory and it works. The detailed description and trigger method for each Skill are introduced on the website.

## Online installation
npx skills add xxx/api-scenario-modeler
npx skills add xxx/api-scenario-runner

# Clone repository
git clone [email protected]:xxx/skills.git

# Copy to Claude Code skills directory
cp -r skills/api-scenario-modeler ~/.claude/skills/
cp -r skills/api-scenario-runner ~/.claude/skills/

# Or copy to WorkBuddy skills directory
cp -r skills/api-scenario-modeler /Users/xxx/.workbuddy/skills/
cp -r skills/api-scenario-runner /Users/xxx/.workbuddy/skills/

Are there any new Skills you want me to make for your work? Leave a comment. The ones with the most likes will be scheduled next month.