verify

name	verify
description	Verify implemented features work correctly by driving browser/UI testing with screenshots and programmatic assertions

Drive interactive testing to verify that implemented features actually work. Take screenshots, record evidence, and enforce programmatic assertions at each step. Uses Playwright MCP when available, falls back to CLI-based verification.

Triggers on: "verify this works", "check the output", "test the UI", "verify the feature", "screenshot test", "visual verification", "does it actually work". Do NOT trigger on: unit test writing (use tdd-workflow), code review (use code-review), standard test runs.

Announce: "Running verification. I will test the feature interactively and capture evidence."

Check for Playwright MCP (`mcp__playwright__*`) or Chrome DevTools MCP (`mcp__chrome-devtools__*`). Fall back to CLI verification if unavailable. From requirements/BDD scenarios, determine what to verify. Each step needs: action, expected result, assertion type. Navigate to feature, interact with UI, capture screenshots at each step, assert expected outcomes. Save screenshots, network logs, console output, assertion results. Pass/fail for each step with evidence links. Every assertion must be programmatic (not visual-only) — check element existence, text content, network responses Capture screenshot BEFORE and AFTER each interaction for comparison Report failures with: step number, expected vs actual, screenshot evidence If Playwright MCP unavailable, verify via CLI commands, API calls, or log output