Agent Browser

Before you start: read this whole install prompt, tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead before you do anything else. Prerequisites — Reinforced framing, disclosed up front: this listing is an instruction file for browser automation through shell commands — at use time the agent drives a REAL browser by running the separate agent-browser CLI (open, snapshot, click, fill, screenshot, pdf), including filling forms and logging into sites with credentials YOU provide in the session. The CLI itself is a separate tool that must already be installed and on PATH: this folder contains only the SKILL.md, no CLI, no installer. Name collision, also disclosed: a different skill also called "agent-browser" (from vercel-labs) is already listed in this catalogue — this one comes from a different source (shanraisshan/claude-code-best-practice) and is not the same listing. No account, key or payment is needed to install this listing itself. Install "Agent Browser" for me. It teaches the agent to automate a real browser through the agent-browser command-line tool: navigate to a URL, take an interactive snapshot whose elements get refs like @e1, then click, fill, select, check, type and scroll by ref, re-snapshotting after every page change — with patterns for form submission, data extraction (including JSON output), screenshots and PDFs, parallel named sessions, headed debugging with recording, local file:// pages, and an iOS Simulator mode (macOS with Xcode and Appium). It also documents saving a logged-in session state to a local file (state save, e.g. auth.json) and loading it back later. Honest scope note, read from the folder itself: the SKILL.md links to six references/*.md deep-dive files and three templates/*.sh scripts that are NOT present in this skill folder in the repository — only the SKILL.md ships here, so those links cannot be followed from this install. Its folder is a single SKILL.md file. Repository: https://github.com/shanraisshan/claude-code-best-practice (the skill is the whole .claude/skills/agent-browser/ folder at the pinned commit: SKILL.md, 1 file in total) Pinned version: install the code at commit cd8e296b06dd1dea732ec15e39645b4edba09bcf exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/shanraisshan/claude-code-best-practice into a temporary folder, check out commit cd8e296b06dd1dea732ec15e39645b4edba09bcf, and copy the .claude/skills/agent-browser/ folder from that checkout. Then compare every file in the folder with the fingerprints recorded for this listing on 2026-10-22: [FINGERPRINTS] - SKILL.md: SHA-256 84effb42371bac499123843088fe4304d8b7c5de765722e3ad92cc0f7cc68557 If a file does not match, it changed since the listing was prepared: do not install it yet, tell me what you found and let me decide. 2. Read SKILL.md in full, and tell me in plain words: what the skill does when used, what it connects to, whether it reads or sends any personal information or credentials, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the files: - The folder holds 1 file only: SKILL.md (7,575 bytes, frontmatter name agent-browser, allowed-tools Bash(agent-browser:*)). There is no script, no CLI binary and no companion file in it — and the references/*.md and templates/*.sh files it links to are absent from this folder in the repository. - Nothing is sent anywhere and no browser is launched during installation itself; the only download at install time is the skill folder from GitHub at the pinned commit. - At use time the skill makes the agent run agent-browser CLI commands in the shell against a real browser: navigating to sites I name, snapshotting pages, clicking and filling forms (including login forms with credentials I provide in the session), extracting page text, and taking screenshots or PDFs. It can also save a logged-in session state to a local file (state save, e.g. auth.json) for reuse. None of that happens during this install, and the CLI itself is not installed by this listing. - Do not run any agent-browser command, do not open any website, and do not use the skill in any way during this install: using it is a separate step that only starts when I ask for it later. 3. When I say "yes, install", install it as a skill: copy the whole agent-browser folder into the agent's skills directory, in a folder named exactly "agent-browser", keeping its 1 file in their original relative layout (SKILL.md at its root). 4. Verify with no network calls: frontmatter valid (name: agent-browser), the 1 file in place with matching fingerprints. Do not run anything from the folder and do not use the skill to verify. 5. Report what was installed, where, and what I still need to do myself (for example: that the agent-browser CLI is a separate tool that must already be installed for this skill to do anything, that a Chromium-based browser may be needed by that CLI, that the iOS mode additionally needs macOS with Xcode and Appium, and that using this skill is a separate step that only starts when I ask for it later). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.