Show HN: ProofShot – Give AI coding agents eyes to verify the UI they build

Summary

Open Source — MIT Licensed Your AI agent says it's done.Get the proof. Let your AI agent prove its own work with video recording, error logs, and proof artifacts — every time it ships code. $ npm install -g proofshot GitHub REC 🧭 Navigate: /board 🖱 Click: + Add ⌨ Type: Enable dark mode ↕ Scroll down 1 🧭 navigate "http://localhost:3000/board" 0:00 2 👁 snapshot "app page" 0:02 3 🖱 click "+ Add" 0:04 4 ⌨ fill "name" "Enable dark mode" 0:06 5 ⌨ fill "owner" "@emma" 0:07 6 🖱 click "P1" priority 0:08 7 🖱 click "Create" 0:10 8 ↕ scroll down 0:12 9 📷 screenshot "page with new item" 0:14 How it works Three commands. Full proof. ProofShot wraps your dev server, records everything the agent does, and outputs a proof bundle when it's done. 1 Start Launches your dev server, opens headless Chromium, and starts recording video. proofshot start --run "npm run dev" 2 Test Your AI agent navigates, clicks, fills forms, and takes screenshots. Every action is logged. proofshot exec navigate "http://localhost:3000" proofshot exec screenshot "homepage" 3 Stop Collects errors, stops recording, trims dead time, and generates your proof artifacts. proofshot stop The viewer Every session, fully reviewable. Video playback synced with an action timeline, screenshots, and element labels — all in a single standalone HTML file. Features Everything you need to verify. Capture, detect, review, and share — all from one CLI. Video Recording Full browser session recording with automatic dead-time trimming. See exactly what the agent did, not the idle waiting. Error Detection Captures browser console errors and scans server logs with pattern matching for JavaScript, Python, Go, Rust, and more. Interactive Timeline A standalone HTML viewer with synchronized video, screenshots, and an interactive action timeline with element labels. PR-Ready Artifacts Generates SUMMARY.md and formatted output ready to paste into pull requests. Visual diff comparison against baselines. Agent-agnostic Works with any AI codi...