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/lftpadilla/agent-dev-kit/web-browsenpx skills add LFTPadilla/agent-dev-kit --skill web-browsegit clone --depth 1 https://github.com/LFTPadilla/agent-dev-kitWhat 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 | $0.00100 | $0.00670 |
| Opus 5 | $0.00050 | $0.00335 |
| Sonnet 5 | $0.00020 | $0.00134 |
| Haiku 4.5 | $0.00010 | $0.00067 |
Grade A, and why
web-browse 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 3d 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 — 65 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Web Browse Skill
Use the browser tool to navigate web pages as a real user.
Mandatory execution rule
Use this skill only when normal brave-search is insufficient: login flows, dynamic pages, forms, buttons, CAPTCHAs, dashboards, or explicit requests to open/navigate a site. For ordinary web search, use brave-search first. Do not tell the user web access is unavailable just because browser interaction is not needed.
Setup
Browser is already configured on Argo:
- Chromium at
/usr/bin/chromium(no sandbox, host mode) - CDP port: 18800
Browser Tool Quick Reference
| Action | When to use |
|---|---|
browser(action="open", url="...") |
Navigate to URL |
browser(action="snapshot", targetId="...") |
Get page structure (aria refs) |
browser(action="screenshot", targetId="...") |
Visual screenshot |
browser(action="act", targetId="...", request={kind:"click", ref:"..."}) |
Click element |
browser(action="act", targetId="...", request={kind:"type", ref:"...", text:"..."}) |
Type in input |
browser(action="act", targetId="...", request={kind:"press", key:"Enter"}) |
Press key |
Workflow
- Open the URL with
browser(action="open", target="host", url="...") - Snapshot to see page structure and get element refs
- Act to interact (click, type, scroll, hover)
- Repeat until task is done
- Close the browser when done
Common Patterns
Login to a site
open → snapshot → find login fields → type credentials → click submit → verify
Extract data from a page
open → snapshot → identify data elements → act to expand/scroll if needed → snapshot → extract
Click through a flow
open → snapshot → click button → wait for navigation → snapshot → click next → ...
Tips
- Use
compact=truein snapshot for fewer nodes - Use
refs="aria"to get stable aria refs for act targets - If a page doesn't load, try adding
timeMs=2000to act for wait - For pages requiring user interaction, use
screenshotto see what the agent sees - Headless mode: add
headless=trueto open if you don't want visible browser - Keep browser open between related actions to avoid re-authenticating
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.
- 3d ago First seen · 65 lines · 100 tokens per session scan A 9c05a60807e1
web-browse is a skill published in the GitHub repository LFTPadilla/agent-dev-kit (2 stars, last pushed 5d ago), licensed MIT. It adds 100 tokens to every session and 670 once invoked, about $0.0005 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 2026-08-31.
Other skills, from other repositories
anysearch
Real-time search engine supporting web search, vertical domain search, parallel batch search, and URL content extraction.
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.
baoyu-url-to-markdown
技术写作员与内容创作者在沉淀网页资料时，当需要抓取包含JS动态渲染或需登录的复杂页面，使用此技能一键将其转换为干净的Markdown格式，自动提取正文与元数据，轻松构建本地知识库，让网页内容归档效率翻倍！.
xiaohongshu-skill
小红书 / Xiaohongshu / RedNote AI Agent Skill。用 Python Playwright 搜索和读取内容、管理登录会话、发布图文/视频/长文、评论、点赞和收藏；默认输出 JSON，任何写操作都必须先获得用户确认。用户提到 xiaohongshu、小红书、rednote、小红书搜索、发到小红书、小红书笔记分析、小红书运营或小红书自动化时触发。.
fanqie-publisher
Publish prepared novel chapters from local Markdown files to the Fanqie Novel writer web backend via browser automation. Use when the user wants to upload chapters, continue publishing unpublished chapters, save or reuse login state, or schedule Fanqie chapter releases from a local directory.
caido-mode
Full Caido SDK integration for Claude Code. Search HTTP history with HTTPQL, test with curl proxied through Caido (caching auth in reusable static curl config files), add match & replace rules, and organize handoffs into named replay sessions and collections - all via the official @caido/sdk-client. PAT auth…