263311487-ux/dsh-verify

README excerpt

dsh-verify Agents self-test and pass. Real browsers tell the truth. dsh-verify is a tiny, dependency-light acceptance tester for agent-delivered web artifacts . You write a JSON spec of what a human would check in a browser; it launches a real headless Chromium, clicks, reads computed styles, and produces an HTML report plus a 0/1 exit code. It exists because we got burned — and we have receipts. The story (why this exists) We ran a 4-agent web team (spec writer → frontend dev → QA → reviewer). The team shipped a demo page with a counter and a dark-mode toggle. Their own review report said: ✅ "All requirements met. No issues found." In a real browser, the toggle button did nothing — the page background never changed. The .dark class was toggled by JavaScript, but the CSS rule for .dark was never written. Every agent self-test passed because there was nothing in the page for the agents to run. No one opened a real browser. That is the gap: agents verify against what they believe they built, not against what a user actually experiences. Unit tests and static checks can't catch a missing CSS rule. This repo contains the bug, fixed, side by side — and the browser-driven evidence that s…

Category