SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsqa-explorer

qa-explorer

Testing & quality

View repository

Updated 10-16 05:46

About this skill

QA Explorer

Skill to intelligently generate test cases — both from documents (pre-dev) and from direct browser exploration (post-dev).

Mode Detection

Determine the mode based on user input:

Input Scenario Mode
User provides a PRD / user story / document / specification text WITHOUT a URL Mode 1 — Pre-Dev
User provides a URL (with or without a document) Mode 2 — Post-Dev
User provides 2 test case files / asks for a "gap report" / "compare" Mode 3 — Gap Only
User provides a URL AND there is already qa-spec-*.json in the folder Mode 2 + auto Gap Report
User provides nothing / ambiguous Ask: "Do you have a PRD document, an application URL, or both?"

Important note: If the user only mentions a feature/flow without a document and without a URL (e.g.: "create test cases for login"), first ask whether there is a document or URL. Do not assume.

Mode 1 — Pre-Dev (Document → @unverified Test Case)

Enable when the input is a document/specification text without a URL.

Read the complete instructions from:

@references/mode-predev.md

Output: File .feature with tag @unverified + file qa-spec-[nama].json.

Mode 2 — Post-Dev (URL → Browser Exploration → @verified Test Case)

Enable when the input contains a URL that can be opened in a browser.

Read the complete instructions from:

@references/mode-postdev.md

Output: File .feature with tag @verified + update/create qa-spec-[nama].json.
Automatic bonus: If there is qa-spec-*.json from Mode 1, automatically add a gap report.

Mode 3 — Gap Only (Compare 2 Test Cases)

Enable when:

  • User asks for a "gap report", "compare test cases", or "which tests are missing"
  • User attaches/mentions 2 different test case files

Read the complete instructions from:

@references/output-format.md#gap-report

Output: File gap-report-[nama]-[tanggal].md.

Default Output Format

All test cases are written in Gherkin (Given-When-Then) format unless the user requests another format.

Supported formats:

Flag Format Output When used
(default) Gherkin / BDD *.feature CI/CD, Cucumber, BDD framework
--format table Markdown Table *.md Review in chat, lightweight documentation
--format xlsx Excel (5 sections) *.xlsx Professional TC management, status tracking, automation reference
--format steps Plain Steps *.md Non-technical teams, simple manual testing
--format playwright Playwright TypeScript *.spec.ts Directly becomes an automation script

Recommendation: Use --format xlsx for the most complete output — covering Test Case Information, Source, Application Overview, Status Summary, and Test Case Table in one Excel file that can be used directly as an official QA document.

Excel Format (--format xlsx) — Complete Structure

The Excel output consists of 5 sequential sections in 1 sheet:

1. TEST CASE INFORMATION  ← metadata dokumen (Feature Name, Created By, dll)
2. SOURCE                 ← asal test case (PRD / URL / dll)
3. APPLICATION OVERVIEW   ← deskripsi singkat aplikasi
4. TEST CASE STATUS       ← ringkasan status (auto-count via COUNTIF)
5. TEST CASE TABLE        ← tabel utama semua test case

Test Case Table Columns: TC ID | Scenario | Summary | Priority | Pre-Conditions | Test Step | Test Data | Expected Result | Actual Result | Test Case Status | Automation Status | Notes

TC ID format: TC0001, TC0002, ... (4 digits, auto-increment)
Priority: P1 - High / P2 - Medium / P3 - Low
Test Case Status: untested (default) / passed / onprogress / blocked / failed / retest
Automation Status: determined based on analysis — not all TCs need automation:

  • automated — script already exists
  • to be automated — automation candidate (repetitive flow, regression)
  • in progress — script is being created
  • manual only — not suitable for automation (visual, external system, rarely run)

See the complete template and guidance for determining status at:

@references/output-format.md#table-format

Browser Automation

This skill supports two browser mechanisms in a hybrid manner:

MCP Playwright (Primary Priority)

If MCP Playwright is available (detected from tool mcp__playwright__* or mcp__executeautomation-playwright-mcp-server__*), use MCP for:

  • Real-time navigation and interaction
  • Direct screenshots
  • Network request inspection
  • Console log monitoring

Bundled Scripts (Fallback)

If MCP is not available, use the scripts in the folder scripts/:

# Install dependencies dulu
cd skills/qa-explorer/scripts && npm install

# Phase 1: Inventarisasi elemen
node scripts/inventory.js --url <URL> [--auth-cookie <cookie>]

# Phase 2 & 3: Happy path + negative tests  
node scripts/happy-path.js --url <URL> --inventory inventory-output.json

# Phase 4: Generate laporan
node scripts/reporter.js --inventory inventory-output.json --results results.json --name <feature-name>

State Management

This skill uses the file qa-spec-[nama].json to store state between sessions:

qa-spec-login.json        ← dari Mode 1 (pre-dev)
login.feature             ← test case @unverified
gap-report-login-2026.md  ← dari Mode 2 atau Mode 3

Automatic detection:

  1. Mode 2 is run → check whether there is qa-spec-*.json in the working directory
  2. If there is → automatically perform gap analysis after exploration
  3. If there is not → save the results as qa-spec-[nama].json new

Naming Convention

Use the feature name as the basis for the file name:

  • qa-spec-[fitur].json (e.g.: qa-spec-checkout.json)
  • [fitur]-unverified.feature (e.g.: checkout-unverified.feature)
  • [fitur]-verified.feature (e.g.: checkout-verified.feature)
  • gap-report-[fitur]-[YYYY-MM-DD].md

Ask the user for the feature name if it is not clear from the context.