AutoBrowse

AutoBrowse is a browser-agent learning loop that improves task-specific behavior without changing model weights. It uses two Claude Agent SDK loops: a doer agent operates the browser, while a teacher agent reviews screenshots, traces, and task results after each run and writes lessons to a strategy.md file for the next attempt. When the strategy converges, the loop produces a reusable skill.md file containing site-specific navigation choices, waiting behavior, and other learned procedures that future agents can apply. The approach depends on replayable environments and verifiable outcomes, and the videos describe monitoring and failure-triggered repair loops for changing websites, with generated skills requiring review because of prompt-injection risks.

browse.sh is a browser-automation platform for AI agents that combines an open catalog of reusable SKILL.md recipes with a command-line interface. The catalog contains website-specific skills for tasks such as searching listings, tracking flights, finding events, and querying public APIs; skills can be installed with the browse CLI, and the videos describe generating new skills on demand by email. The CLI also provides low-level browser actions such as clicking, typing, selecting, scrolling, hovering, pressing keys, and taking screenshots, along with network and console tracing for debugging. It can run commands against local Chromium or remote Browserbase sessions, including cloud search and fetch operations.

BrowserSkill is an open-source browser-automation bridge for AI agents, consisting of the `bsk` CLI and daemon plus a Chrome or Microsoft Edge extension. An agent sends browser tasks through the CLI; the local daemon routes them over WebSocket to the extension, which performs actions in a separate visible Agent Window. Agents can use existing browser login state, borrow an already open tab only with explicit approval, and request human intervention for captchas, confirmations, or other user-only steps. It supports shell-capable agent harnesses, macOS, Linux, and Windows, and is distributed under the MIT license.

An installable browser-automation CLI for AI agents, hosted by browse.sh. It provides an open catalog of reusable SKILL.md recipes for website-specific tasks, which agents can install on demand with commands such as `browse skills add alltrails.com`. It also exposes low-level browser primitives for clicking, scrolling, typing, hovering, pressing keys, selecting elements, and taking screenshots, addressing elements by DOM selectors or accessibility references. The CLI can tail browser network and console output for debugging and can run commands against local Chromium or remote Browserbase sessions. Catalog skills cover website operations such as searching listings, tracking flights, finding events, and querying public APIs; the page states that suggested DOM selectors and XHR requests can reduce token usage.

A GitHub-hosted hands-on workshop repository for wiring AI-powered UI verification workflows. It covers a podcast landing-page challenge with easy and medium paths, using a Chrome extension for deterministic visual annotation, a coding agent, and the Nova Act SDK to find and fix rendered UI discrepancies and verify design expectations and user flows. The annotation workflow lets participants select elements, draw on a page, or drop pins and export structured feedback; the verification workflow crawls a site, checks the resulting interface, and generates a report. The repository includes workshop instructions, setup prerequisites, integration code, and a Bee Wearable reverse proxy that surfaces completed-conversation feedback as sidebar cards with Apply and Dismiss actions.