enixcoda/replayright

Replayright is a tool that helps AI coding agents write browser automation tests without the usual back-and-forth of writing, running, fixing, and rewriting code. The typical workflow for an agent creating an end-to-end test (a test that controls a real browser and simulates user actions) involves exploring one step, manually writing code for it, running the whole thing to check it works, then repeating for the next step. Replayright removes that loop. It works by acting as an "agent skill", a plugin that extends what an AI coding agent can do. When the agent explores a website and performs actions (opening a URL, filling in a form field, clicking a button), Replayright records each of those actions automatically in the background. By the time the agent finishes exploring, there is already a complete, replayable script saved to disk. No separate write step is needed. The code examples in the README show this clearly. An agent given a single instruction to fill out a web form and submit it issues commands like open, fill, click, and close. After the close command, Replayright saves the recorded script as a JavaScript file. The same flow works interactively, where a user guides the agent step by step and saves the script at the end. Replayright follows the agent skill standard from agentskills.io, which means it is designed to work with any compatible AI coding agent. The project is written in JavaScript and is licensed under MIT. You would use it when you want an AI agent to produce a reliable browser test in a single pass rather than through repeated write-run-debug cycles.