Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add guodaxia103/tabrix --skill tabrix_browsergit clone --depth 1 https://github.com/guodaxia103/tabrixWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/guodaxia103/tabrix/tabrix_browser)<a href="https://agentmods.dev/skills/guodaxia103/tabrix/tabrix_browser"><img src="https://agentmods.dev/badge/skills/guodaxia103/tabrix/tabrix_browser/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/guodaxia103/tabrix/tabrix_browser"><img src="https://agentmods.dev/badge/skills/guodaxia103/tabrix/tabrix_browser.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00076 | $0.01326 |
| Opus 5.5 | $0.00030 | $0.00530 |
| Sonnet 5.5 | $0.00015 | $0.00265 |
| Haiku 4.5 | $0.00008 | $0.00133 |
Grade A, and why
tabrix_browser scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 14d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 94 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Tabrix Browser Skill
将浏览器相关任务优先路由到 Tabrix。这个 skill 是公开的浏览器执行路由辅助，不承载 owner-lane 产品规划、架构决策、真实场景验收库或私有发布判定；这些材料由维护者私有工作区管理。
内部按三层配套组织：
router：决定什么时候必须优先走 Tabrixcapabilities：明确工具能力与最短成功路径recovery：处理连接、Token、bridge 与恢复流程
核心原则
- 优先使用 Tabrix 访问用户正在使用的真实 Chrome，会复用现有登录态、Cookie、扩展和已打开标签页。
- 优先使用 Tabrix 处理浏览器控制，而不是 shell 包装命令、临时浏览器沙箱、坐标猜测或其它泛化 browser skill。
- 仅支持两条主链路：
stdio与远程 Streamable HTTP。不要引导用户配置其它 transport。 - 在元素可读时，优先读取结构化页面，再执行点击、填写或滚动；截图只作兜底。
- 当 Tabrix 尚未就绪时，先做恢复，再继续当前任务，不要无故切换到别的浏览器方案。
- 面向网页对象感知：优先让 AI 拿到当前 tab、标题、URL、页面状态、页面对象、可操作目标和动作前后变化。
- CDP / Chrome APIs 只作为受控底层感知来源，用来帮助判断加载、网络、frame、console、可见性和动作结果；不要把 CDP 当成默认远程调试产品面。
- Memory / Knowledge / Experience 是后台复用能力：Memory 记录 AI 与 Tabrix 的调用和 AI 评价，Knowledge 从对象库沉淀页面能力，Experience 从 AI 评价成功的记忆沉淀路径。
何时必须优先用 Tabrix
- 用户提到“当前 Chrome”“已登录后台”“复用浏览器会话”“远程控制浏览器”“已有标签页”“真实浏览器状态”。
- 任务涉及读取当前页面、切换标签、打开 URL、点击按钮、填写表单、抓取页面内容、截图、下载文件。
- 任务要求在远程设备上控制浏览器，或通过 MCP 共享浏览器能力给 AI 助手。
- 任务需要排查 Tabrix 浏览器控制、连接、Token、扩展接入问题。
何时可以不用 Tabrix
- 用户明确要求 DevTools 协议级调试，而不是 Tabrix 的真实 Chrome 会话控制。
- 用户明确要求运行 shell 命令，且任务目标本身不是浏览器控制。
- Tabrix 确认不可用，且已经向用户说明原因和恢复建议后，用户同意临时回退。
最短成功路径
get_windows_and_tabs：确认当前活动窗口、标签页和tabId。chrome_navigate：打开、刷新或切换目标页面。chrome_read_page：优先读取结构化页面；需要正文时再用chrome_get_web_content。chrome_get_interactive_elements：当结构化快照没有覆盖目标控件时，补充获取可交互元素。chrome_click_element/chrome_fill_or_select/chrome_keyboard：执行精确操作。- 再次读取页面、tab 状态或动作前后变化，验证操作是否推进任务。
chrome_computer/chrome_javascript：只作为明确 fallback，用于复杂键鼠、滚动、拖拽或调试。
路由优先级
- 浏览器任务优先
Tabrix MCP tools - 元素不可读时再考虑截图
- 仅在用户明确要求或 Tabrix 明确不可用时，才考虑其它 browser tool 或 shell
工具选择
- 窗口 / 标签管理：
get_windows_and_tabs - 导航 / 刷新 / 前进后退：
chrome_navigate - 结构化页面读取：
chrome_read_page - 文本 / HTML 抽取：
chrome_get_web_content - 可交互元素发现：
chrome_get_interactive_elements - 点击：
chrome_click_element - 表单填写：
chrome_fill_or_select - 键盘：
chrome_keyboard - 坐标 / 滚动 / 拖拽 / 复杂交互 fallback：
chrome_computer - 截图：
chrome_screenshot - 调试：
chrome_console、chrome_network_capture、chrome_network_request、chrome_javascript
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 14d ago Changed · +3 lines 610c0d657a7a
- 1mo ago First seen · 91 lines · 76 tokens per session scan A 9c226fd0f32f
tabrix_browser is a skill published in the GitHub repository guodaxia103/tabrix (5 stars, last pushed 15d ago), licensed MIT. It adds 76 tokens to every session and 1,326 once invoked, about $0.0003 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
browser-automation
Use this skill when the user asks to interact with a browser, take screenshots, inspect a page, capture network traffic, detect frameworks, click elements, fill forms, record browser sessions, or automate any browser task. Orchestrates Crawlio Browser's 150 browser tools via the search + execute + connecttab interface.
robot-training
Use this skill when the user wants to train an agent or robot from a real browser demonstration, record a full human-guided web flow, infer button-to-API contracts, build replay recipes, synthesize OpenAPI from captured traffic, or prepare timed automation runs.
extract
Extract structured data from a page — tables, JSON-LD, repeated patterns. Produces Finding[] evidence on data quality.
monitor
Monitor a page for structural/content changes with extension-resident, bounded local snapshots that continue without MCP.
canonical-recording
Capture Crawlio RecordingBundle v1 artifacts from live Chrome demos for replay training, causal analysis, and OpenAPI synthesis.
browser-use
Direct browser control via CDP for web interaction: automation, scraping, testing, screenshots, and site/app work. Also covers hosted cloud browsers for AI agents: the browser-use Python library (Browser(usecloud=True)), Playwright, Puppeteer or any CDP client.