| name | develop-test-mcp |
| description | Develop a Playwright automated test using Playwright MCP browser tools for live element exploration. Use when the user provides a test case, asks to automate a scenario, or says "write a test for X". Requires Playwright MCP server to be running (configured in .mcp.json).
|
| allowed-tools | Read, Grep, Glob, Write, Edit, Bash(pnpm:*), Bash(npx playwright:*), Bash(git:*), mcp__playwright__browser_navigate, mcp__playwright__browser_snapshot, mcp__playwright__browser_click, mcp__playwright__browser_fill, mcp__playwright__browser_select_option, mcp__playwright__browser_evaluate, mcp__playwright__browser_screenshot, mcp__playwright__browser_wait_for, mcp__playwright__browser_type, mcp__playwright__browser_hover, mcp__playwright__browser_close |
Develop Test — MCP Edition
🔴 Hard Prohibitions
- NEVER read
.tpl Smarty template files — use MCP snapshot / evaluate on the live page.
- NEVER navigate directly to deep pages via constructed URLs — walk the full UI flow from the homepage or login page.
Step 1 — Analysis
Goal: Understand intent, find what already exists, avoid duplication.
Read docs-mcp/maps/page-object-map.md:
- Pick the correct spec file and fixture from the Scenario → Spec File table.
- Scan existing PO methods and locators — extend, never duplicate.
Identify from the scenario: pages involved · actions · assertions.
Duplicate Gate: Search for a test with the exact scenario name. If found, report the file path and ask the user how to proceed — do NOT continue silently.
