Agentic Testcase Generator

AI-powered test case generation from requirement documents. Works with Claude Code, Cursor AI, GitHub Copilot, OpenAI Codex, Windsurf, and Google Antigravity.

Drop your PRD/SRS/spec into input/, and the AI agent walks you through a structured 3-step workflow: Refinement → Planning → Execution — with human review at every step.

Supported Platforms

| Platform | Config format | |----------|--------------| | Claude Code | .claude/ (agents, skills, rules) | | Cursor AI | .cursor/rules/*.mdc | | GitHub Copilot | .github/copilot-instructions.md | | OpenAI Codex | AGENTS.md | | Windsurf (Codeium) | .windsurfrules | | Google Antigravity | AGENTS.md |

Quick Start

Option 1: npx (recommended)

# Setup for your AI tool
npx agentic-testcase-generator init --platform claude
npx agentic-testcase-generator init --platform cursor --lang vi
npx agentic-testcase-generator init --platform claude,cursor,copilot

Option 2: Clone from GitHub

git clone https://github.com/ngochiep30051998/agentic-testcase-generator.git
cd agentic-testcase-generator
npm install

# Then init for your platform
node bin/cli.js init --platform claude --lang vi

Option 3: npx clone (for customization)

npx agentic-testcase-generator clone --platform cursor --lang vi

CLI Commands

# Scaffold config for a platform
npx agentic-testcase-generator init --platform <name> [--lang vi|en]

# Clone full repo for customization
npx agentic-testcase-generator clone [--dir <path>] [--platform <name>] [--lang vi|en]

# List supported platforms
npx agentic-testcase-generator platforms

| Option | Description | |--------|-------------| | --platform | Target platform: claude, cursor, copilot, codex, windsurf, antigravity. Multiple: claude,cursor | | --lang | Skill language: en (default) or vi (Vietnamese) | | --dir | Clone destination (default: agentic-testcase-generator) |

Init Examples by Platform

# Claude Code
npx agentic-testcase-generator init --platform claude --lang vi
# → Creates: .claude/agents/, .claude/skills/, .claude/rules/, CLAUDE.md

# Cursor AI
npx agentic-testcase-generator init --platform cursor --lang vi
# → Creates: .cursor/rules/testcase-generator.mdc

# GitHub Copilot
npx agentic-testcase-generator init --platform copilot --lang en
# → Creates: .github/copilot-instructions.md

# OpenAI Codex
npx agentic-testcase-generator init --platform codex --lang en
# → Creates: AGENTS.md

# Windsurf (Codeium)
npx agentic-testcase-generator init --platform windsurf --lang vi
# → Creates: .windsurfrules

# Google Antigravity
npx agentic-testcase-generator init --platform antigravity --lang vi
# → Creates: AGENTS.md

# Multiple platforms at once
npx agentic-testcase-generator init --platform claude,cursor,copilot --lang vi

How It Works

1. Place documents in input/

| Format | Support | |--------|---------| | .md (Markdown) | Recommended | | .txt (Text) | Direct read | | .pdf | Max 20 pages/read | | .doc / .docx (Word) | Auto convert to .md |

2. Ask your AI tool to generate

# Claude Code
claude "Generate testcases from input/my-feature.md"

# Cursor / Copilot / Others
Open your AI chat and type: "Generate testcases from input/my-feature.md"

3. AI walks you through 3 steps

Step 0 (Optional): Training

• Place sample testcases in training/
• Say "training" to activate
• AI learns writing style, proposes new usecases
• You approve → say "next"

Step 1: Refinement

• AI reads document, extracts features, entities, fields, roles
• Suggests areas needing more coverage
• Asks clarifying questions
• You answer questions → say "next"

Step 2: Planning

• AI classifies tests by module, proposes Excel sheet structure
• Estimates TC count based on actual complexity (not limited by config)
• Presents detailed plan: modules, features, estimated TCs
• You adjust if needed → say "generate"

Step 3: Execution

• AI asks: merge into one file or split per module?
• AI asks: sequential or parallel writing? (if >= 3 modules)
• Writes test cases, exports Markdown + Excel
• You review → AI iterates on feedback

4. Get results in output/

| File | Description | |------|-------------| | *-testcases.md | Test cases in Markdown | | *-testcases.xlsx | Excel with multiple sheets + Summary sheet | | *-summary.md | Summary report |

Excel includes:

• Summary sheet: Total TCs, Pass/Fail/Blocked/Skip, Pass Rate
• Module sheets: Separated by module, e.g. "Login (8 TCs)"
• Columns: No, ID, Type, Description, Steps, Expected Result, Priority, Dev, QC, Note

Configuration

Edit config.yaml:

Output language

language: vi    # vi (Vietnamese) | en (English)

Coverage depth

coverage_depth: standard

| Value | Description | TCs/feature (reference) | |-------|-------------|------------------------| | minimal | Happy path + basic validation | ~10-15 | | standard | All main categories | ~20-40 | | thorough | Deep edge cases, combinatorial | ~40-80+ |

TC count is a suggestion. AI evaluates actual complexity and proposes accordingly.

Test case categories

categories:
  functional: true        # Happy path, CRUD
  validation: true        # Required fields, format, length
  boundary: true          # Min/max, edge values
  negative: true          # Invalid data, error states
  ui_ux: true             # Loading, disabled states
  authorization: true     # Role-based access
  error_handling: true    # Network errors, timeouts
  empty_state: true       # No data states
  integration: false      # Cross-module
  e2e: false              # Full user journeys
  security: false         # XSS, injection, CSRF
  performance: false      # Load times, limits

Project info (optional)

project:
  name: "My Project"
  type: "web"             # web | api | mobile
  stack: "NestJS + React"
  base_url: "http://localhost:3000"
  api_url: "http://localhost:8080/api/v1"

Utilities

# Export Excel manually
node scripts/export-xlsx.js output/feature-testcases.md
node scripts/export-xlsx.js --lang en output/feature-testcases.md

# Convert Word to Markdown
node scripts/convert-word.js input/document.docx
node scripts/convert-word.js input/   # Convert all .doc/.docx

Project Structure

agentic-testcase-generator/
├── bin/
│   └── cli.js                # CLI entry point
├── core/                     # Skills in English (source of truth)
│   ├── skills/               # 0-training, 1-refinement, 2-planning, 3-execution
│   ├── rules/                # output-format, coverage-standards
│   ├── templates/            # Markdown + summary templates
│   └── config.yaml           # Default config
├── i18n/
│   └── vi/                   # Vietnamese translation
│       ├── skills/
│       ├── rules/
│       └── templates/
├── adapters/                 # Platform-specific generators
│   ├── claude.js
│   ├── cursor.js
│   ├── copilot.js
│   ├── codex.js
│   ├── windsurf.js
│   └── antigravity.js
├── scripts/                  # Utilities
│   ├── export-xlsx.js        # Export Excel (.xlsx)
│   ├── export-csv.js         # Export CSV (UTF-8 BOM)
│   └── convert-word.js       # Convert Word to Markdown
├── templates/                # Output templates
├── training/                 # Sample documents for training
├── input/                    # Requirement documents go here
├── output/                   # Generated test cases
├── config.yaml               # User configuration
├── package.json
└── README.md

Author

Nguyen Ngoc Hiep

• Website: hiepnn.com
• LinkedIn: linkedin.com/in/hiep-nguyen

License

MIT License — Free and open source. Use, modify, and distribute freely.

If you find this useful, please give it a Star on GitHub!