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 agentmods add skills/seed-forge/harness-ai-kit/devlab-web-visual-opsnpx skills add seed-forge/harness-ai-kit --skill devlab-web-visual-opsgit clone --depth 1 https://github.com/seed-forge/harness-ai-kitWrote 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/seed-forge/harness-ai-kit/devlab-web-visual-ops)<a href="https://agentmods.dev/skills/seed-forge/harness-ai-kit/devlab-web-visual-ops"><img src="https://agentmods.dev/badge/skills/seed-forge/harness-ai-kit/devlab-web-visual-ops.svg" alt="Measured on agentmods" 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 | $0.00176 | $0.03071 |
| Opus 5 | $0.00088 | $0.01536 |
| Sonnet 5 | $0.00035 | $0.00614 |
| Haiku 4.5 | $0.00018 | $0.00307 |
Grade A, and why
devlab-web-visual-ops 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 4d 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 — 162 lines — stays where its author put it; the contents beside it link to each section on GitHub.
DevLab Web Visual Ops — Web 视觉化操作
Purpose
对 Web 应用做可重复的视觉化操作:观察页面、导航、交互、检查(Console/Network/Runtime)、验证、截图捕获,并沉淀为 Debug / UI Review / Visual Regression / E2E / Browser Automation 五类 Workflow 的标准化产物。
定位边界(与 Stagehand 的关系):Stagehand 本质是 AI Browser Automation;本技能是范围更大的 Web Visual Operations。因此 Stagehand 是 Router 层的 AI Adapter 之一,不是 Skill 底座。Console/Network/Trace/截图/视觉回归/UI Review 是 Skill 自己的核心能力,不抽象成"Stagehand 能力",避免被具体框架绑死。
适用条件
- 目标是有浏览器 UI 的 Web 应用(SPA / 微前端 / MPA 均可);
- 需要浏览器端到端执行的操作:视觉走查、UI 审查、页面调试、截图对比、视觉回归、浏览器自动化;
- 执行环境可运行 Node ≥ 18 + Playwright。浏览器获取方式可选(三选一):①
npx playwright install chromium自动下载到 ms-playwright 缓存;② 复用已有 ms-playwright 缓存;③ 系统 Chrome/Edge(executablePath或channel)。无头模式默认可用:Playwright 与 Vision/CUA 路径还可用更轻的chromium_headless_shell(实测三项全 PASS)。 - Stagehand AI 增强路径需 Node ≥ 22.18.0 + newapi 消费 key(
chat_default模型,见 infra-aimodel-ops);必须完整 Chromium(headless 模式即可,实测全 PASS),headless shell 不支持扩展注入(实测失败); - 无 DOM 或 DOM 不可靠的场景(canvas / 复杂 iframe / 纯视觉渲染)走 Vision/CUA 兜底。
核心设计:三层架构
devlab-web-visual-ops
│
├── Core 能力原语(Observe / Navigate / Interact / Inspect / Validate / Capture)
├── Router 执行路由(Playwright / Stagehand / DevTools / Vision-CUA)
└── Workflows 业务闭环(Debug / UI Review / Visual Regression / E2E / Browser Automation)
- Core 定义"做什么",与具体执行器解耦;所有 Workflow 只调用 Core 原语。
- Router 决定"用谁执行":默认确定性执行(Playwright);AI 增强/非确定性页面切 Stagehand;Console/Network/Runtime 检查走 DevTools;无 DOM 走 Vision/CUA。统一输入
task + page,统一输出action / observation / extraction协议,保证执行器可替换。 - Workflows 组装"业务闭环":每个 Workflow 是 Core 原语 + Router 选路的固定编排,输出标准化产物。
资产结构
.visual-ops/
├── artifacts/ # 产物:screenshots/ traces/ hars/ 报告/ 视频
│ ├── baselines/ # 视觉回归基线(更新需用户确认 HD-3)
│ └── reports/ # Debug / UI Review / Regression 报告
├── cases/ # E2E / 自动化用例(复用 devlab-web-test-e2e 约定)
└── config.local.* # 项目侧本地覆盖(不入库)
工作流
0. 入口:意图解析与 Router 判定
- 读取
~/.harness-ai-kit/config.yaml的assets.devlab-web-visual-ops段(preview_base_url / preview_start_command / stagehand_enabled 等);缺失时按项目探测。 - 按用户意图归类到 Workflow(Debug / UI Review / Visual Regression / E2E / Browser Automation);意图不明确时询问用户。
- 按下表选择 Router 执行器;执行器不可用时按降级路径切换并记录原因。
What ships with it
9 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.
- 4d ago First seen · 162 lines · 176 tokens per session scan A 91dcb4dbebbd
devlab-web-visual-ops is a skill published in the GitHub repository seed-forge/harness-ai-kit (21 stars, last pushed 7d ago), licensed Apache-2.0. It adds 176 tokens to every session and 3,071 once invoked, about $0.0009 per session on Opus 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 2026-08-30.
Other skills, from other repositories
9router-web-fetch
Fetch URL → markdown / text / HTML via 9Router /v1/web/fetch using Ollama Cloud / Firecrawl / Jina Reader / Tavily Extract / Exa Contents. Use when the user wants to scrape a webpage, extract URL content, read article, or convert a URL to markdown.
browser-testing-with-devtools
Tests in real browsers via Chrome DevTools MCP. Use when building or debugging anything that runs in a browser. Use when you need to inspect the DOM, capture console errors, analyze network requests, profile performance, or verify visual output with real runtime data. Requires the chrome-devtools MCP server to be…
playwright-recording
Record browser interactions as video using Playwright. Use for capturing demo videos, app walkthroughs, and UI flows for Remotion videos. Triggers include recording a demo, capturing browser video, screen recording a website, or creating walkthrough footage.
competitor-ad-intelligence
Scrape competitor ads from Meta, TikTok, Google, and LinkedIn ad libraries, analyze creative patterns (hooks, formats, CTAs), reverse-engineer landing page funnels, and produce a strategic teardown with vulnerability analysis and counter-play recommendations. Use when you need to understand the competitive ad…
qa/e2e-playwright
Playwright E2E 测试完整方法论,涵盖项目初始化、Page Object Model、认证复用、API Mock、视觉回归、多浏览器测试、CI 集成和调试技巧.
render-chatgpt-chat
Assemble a ChatGPT chat-reveal video ad from a thread + timeline JSON — one continuous Playwright recording of a ChatGPT mobile chat (user types with the iOS keyboard up → taps send → keyboard slides down + header cluster swaps in one beat → one gray loading dot → the assistant answer streams in word-by-word)…