/generate

Communication Style: Casual, professional Bahasa Indonesia.

Mandatory Rules

1. If the user has not provided a path folder or path file input, always ask with options (option list).
2. If the user has provided a valid path, do not ask again.
3. Do not generate without a test plan markdown.
4. If there is a conflict, always prioritize the user's instructions.

Instructions

Think to execute step by step :

A. Determine Operation Mode

1. Determine whether the user wants:

• Mode A: Generate from folder → input: test plan folder path
• Mode B: Generate from file → input: single test plan file

2. If the user provides only one path, automatically detect whether that path is a folder or a file.
3. If the path is not valid → stop the process and notify the user.

B. Collect Test Plan Markdown

Mode A (Folder)

1. Scan folder: docs/test-plan/<feature>/**.md
2. Sort by scenario ID (SCN-001 … SCN-999)

Mode B (File)

1. Take only 1 test plan markdown file.
2. If no valid markdown is found, stop.

C. Generate Playwright Test Code

1. Ensure the required ENV variables are available in tests/.env or tests/.env.example. Minimum: BASE_URL, AUTH_EMAIL, AUTH_PASSWORD.
2. Make sure the application is running according to the BASE_URL env. Bash(curl -s -o /dev/null -w "%{http_code}" BASE_URL || echo "Application not running")
3. Eksekusi npm run dev jika "Application not running".
4. Read all plan markdown from step B.
5. Create a TODOS list to generate test file based on Test Plan Markdown files.
6. Activate skill test-file

7. Generate all test file and make sure to follow test-file skill.
8. Deletage to playwright-qa-fixer agent to run the generated test and fix it if there are any failed tests

D. Quality Check

1. Detect linting tooling: If there is a lint script in package.json or eslint is installed, assume linting is allowed by default.
2. If eslint is available:
   • Run lint + auto-fix limited to the newly created test files, e.g., npx eslint tests/<feature_name>/<scenario-code>-<kebab-case-scenario>.spec.ts --fix
   • If lint fails due to configuration, display a brief error and continue.
3. If no linting tooling is available: Briefly explain that lint was not run because no configuration was detected.

E. Closing

1. After generation is complete, close the session with: "Test file saved at <generated_path> and is ready to be run."