What problem does it solve? Agents often claim a fix works by reading code and guessing runtime behavior, producing unverified conclusions. This Skill replaces inference with first-hand evidence: real browser rendering, structured logs, metrics, and traces, so every verification result is backed by machine-checkable proof. ## Core Features & Use Cases - Browser-Driven Verification: Reproduce UI bugs, validate user journeys, and capture before/after screenshots and DOM snapshots using Playwright or Puppeteer, including mobile emulation and network throttling. - Observability Verification: Translate performance and reliability constraints (e.g., P99 latency < 800ms, error rate < 0.1%) into queryable assertions against structured logs, Prometheus metrics, and distributed traces. - Evidence-Backed Acceptance: Enforce machine-checkable acceptance criteria, attach timestamped screenshots and query results to PRs or execution plans, and report capability gaps when tooling is missing. - Use Case: A user reports the login page shows a white screen. The agent drives the page in a headless browser, captures a screenshot with timestamp and URL, finds a console JS error, and reports the evidence to the implementation agent. ## Quick Start Ask the agent to verify that the checkout flow fix actually works by driving the full user journey in a browser and attaching before-and-after screenshots plus API latency metrics as evidence.
