web-access
Skill4 starsChanged 2 months agoMITView on GitHub
PaperPilot 的底层联网能力层。所有联网操作(搜索、网页抓取、文件下载、登录后操作)统一由此 skill 调度。 各阶段 skill(选题/文献/数据/写作)通过声明搜索意图调用本 skill,本 skill 负责通道选择和实际执行。
- Pipes a download into a shell
- Sends data out
22 other projects use a version of this.
Adopted by
Most recently taken up by
- cjh-solver/MathModeling-Skills-Arsenaltheir own version, first seen Nov 8, 2026
- smiling66652/workbuddy-skill-toolkitstheir own version, first seen Nov 1, 2026
- hua695/MathModeling-Skills-Arsenaltheir own version, first seen Oct 25, 2026
- superdesigndev/tregtheir own version, first seen Oct 17, 2026
- eze-is/web-accesstheir own version, first seen Oct 14, 2026
- leishifu666/skillstheir own version, first seen Oct 9, 2026
Used with
Projects with it also use markitdown-skill, humanizer-zh and 4 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
Glen-WeiSkillmarkitdown-skill
Convert documents to Markdown using Microsoft's MarkItDown CLI (`markitdown`). Supports PDF, Word, PowerPoint, Excel, images (OCR), audio (transcription), HTML, YouTube, and URLs. Use when a user wants to convert a file or URL to Markdown, extract text from PDFs/images, transcribe audio/video, or batch-convert documents.
Also in 3 projects that use this
huangservaSkillhumanizer-zh
Also in 3 projects that use this
AnthropicSkillpdf
Comprehensive PDF manipulation toolkit for extracting text and tables, creating new PDFs, merging/splitting documents, and handling forms. When Claude needs to fill in a PDF form or programmatically process, generate, or analyze PDF documents at scale.
Also in 4 projects that use this
AnthropicSkillpptx
Use this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates (.potx), layouts, speaker notes, or comments. Trigger whenever the user mentions \"deck,\" \"slides,\" \"presentation,\" or references a .pptx or .potx filename, regardless of what they plan to do with the content afterward. If a .pptx or .potx file needs to be opened, created, or touched, use this skill.
Also in 3 projects that use this
OpenAISkillbrainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.
Also in 3 projects that use this
- +1 more from anthropics
Other versions22 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 22
- eze-is/web-access★ 9.3ksame name
- superdesigndev/treg★ 7.7ksame name
- binggandata/bggg-skills★ 601same name
- Ce-Legend/ce-legend-skills★ 55same name
- SoFarSoGoodya/math-history-paper★ 17same name
- cjh-solver/MathModeling-Skills-Arsenal★ 14same name
- hua695/MathModeling-Skills-Arsenal★ 12same name
- b1mango/Agent-Skills★ 4same name
- finger88/my-claude-skills★ 3same name
- bushuhui/skills★ 2same name
12 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- paperpilotSkill
- data-collectorSkill
- integrity-auditorSkill
- literature-reviewSkill
- paper-searchSkill
- paper-writerSkill
- topic-explorerSkill
The file
Files
Web Access Skill
基于 eze-is/web-access v2.5.3 (MIT) 适配为 PaperPilot 内置底层能力层,兼容所有 Agent 平台
定位
各阶段 skill(选题/文献/数据/分析/写作)
│ 声明:搜索意图 + 判断标准 + 输出格式
▼
web-access(本 skill — 底层能力层)
│ 执行:通道选择 + 实际获取 + 文件下载 + 站点经验
▼
工具层(WebSearch / WebFetch / curl / paper-search-mcp / CDP)职责边界:
- 本 skill 负责"怎么联网"——通道选择、执行策略、站点经验、失败处理
- 上游 skill 负责"搜什么"——搜索维度、关键词、结果判断标准、输出格式
前置检查
在开始联网操作前,检查可用能力层级:
| 层级 | 能力 | 依赖 |
| 1(基础) | WebSearch + WebFetch + curl 下载 | 无(Agent 内置) |
| 2(增强) | + paper-search-mcp + tavily | MCP 配置 |
| 3(完整) | + CDP 浏览器自动化 | Node.js 22+ + 浏览器调试 |
CDP 模式检查:
node "skills/web-access/scripts/check-deps.mjs"exit 0→ CDP 可用exit 2→ 需询问用户浏览器偏好exit 1→ CDP 不可用,降级到层级 1-2
层级 1-2 足够完成绝大多数研究任务。 CDP 主要用于需要登录态或反爬严格的站点。
浏览哲学
像人一样思考,兼顾高效与适应性地完成任务。
① 拿到请求 — 明确成功标准:什么算完成了?
② 选择起点 — 根据任务性质选最可能直达的方式验证。
③ 过程校验 — 每步结果都是证据。方向错了立即调整,不反复重试同一方式。
④ 完成判断 — 对照成功标准确认完成,不过度操作。
联网工具选择
核心原则:搜索不是终点,获取到文件/数据才是。一手信息优于二手信息。
多平台工具映射
| 能力 | Antigravity (AGY) | 通用/其他 Agent |
| WebSearch | search_web | WebSearch / 内置搜索 |
| WebFetch | read_url_content | WebFetch / 内置 fetch |
| 文件下载 | run_command (curl -L -o) | bash + curl |
| 学术搜索 | paper-search-mcp / literature-search-openalex | paper-search-mcp |
| 浏览器自动化 | read_browser_page | CDP Proxy API |
场景→通道决策
| 场景 | 工具 |
| 搜索摘要或关键词结果,发现信息来源 | WebSearch |
| URL 已知,定向提取特定信息 | WebFetch |
| URL 已知,需要原始 HTML(meta、JSON-LD) | curl |
| 下载文件(PDF/Excel/CSV/数据集) | curl -L -o <path> <url> |
| 学术论文搜索/元数据/PDF下载 | paper-search-mcp |
| 非公开内容 / 反爬严格的平台 | 浏览器 CDP |
| 需要登录态、交互操作 | 浏览器 CDP |
| 动态页面 / JS 渲染 | WebFetch(rendered) → CDP |
Jina(可选,节省 token):r.jina.ai/example.com,网页转 Markdown。适合文章/文档;不适合数据面板。限 20 RPM。
文件下载规范
搜到 URL 后必须实际下载,不能止步于列清单。
# 下载
curl -L -H "User-Agent: Mozilla/5.0" -o "<目标路径>" "<url>"
# 验证
file <downloaded_file> # 确认类型
wc -c <downloaded_file> # 确认非零失败处理:
- 403/登录墙 → 标记"需用户手动获取",给具体步骤
- 0字节/HTML内容(应为PDF) → 换来源或升级到 CDP
- 超时 → 重试一次
阶段策略扩展点
各阶段 skill 通过以下模式声明搜索需求,web-access 负责执行:
策略声明格式
### web-access 搜索策略
**搜索意图**: [学术文献 / 数据源定位 / 信息侦察 / 文件获取]
**成功标准**: [什么算完成]
**优先源**: [paper-search-mcp / 巨潮资讯网 / 政府官网 / ...]
**输出要求**: [下载到data/raw/ / 提取文本 / 结构化为CSV / ...]
**失败降级**: [换什么源 / 标记为手动]各阶段典型策略
| 阶段 | 搜索意图 | 首选通道 | 成功标准 |
| 选题探索 | 信息侦察(验证方向) | WebSearch + WebFetch | 获得方向判断依据 |
| 文献调研 | 学术文献检索 + PDF下载 | paper-search-mcp + curl | 论文元数据+PDF存入raw/ |
| 数据搜集 | 数据源定位 + 文件下载 | WebSearch → WebFetch → curl | 数据文件存入data/raw/ |
| 实证分析 | 方法论参考 | paper-search-mcp + WebFetch | 找到模型设定参考 |
| 论文写作 | 规范查询(期刊格式) | WebFetch 直达 | 获取格式要求 |
浏览器 CDP 模式
通过 CDP Proxy 直连用户日常浏览器(Chrome / Edge),天然携带登录态。
启动
node "skills/web-access/scripts/check-deps.mjs"Proxy API(localhost:3456)
curl -s http://localhost:3456/targets # 列出 tab
curl -s -X POST --data-raw '<url>' http://localhost:3456/new # 新建 tab
curl -s -X POST "http://localhost:3456/eval?target=ID" -d '<js>' # 执行 JS
curl -s "http://localhost:3456/screenshot?target=ID&file=<path>" # 截图
curl -s -X POST "http://localhost:3456/click?target=ID" -d '<sel>' # 点击
curl -s "http://localhost:3456/scroll?target=ID&direction=bottom" # 滚动
curl -s -X POST --data-raw '<url>' "http://localhost:3456/navigate?target=ID" # 导航
curl -s "http://localhost:3456/close?target=ID" # 关闭 tab何时用 CDP
- 需要登录态(巨潮资讯网会员区、知网全文等)
- 反爬严格(动态加载、需交互才显示内容)
- WebFetch rendered 模式仍失败
- 需要填表、翻页等交互操作
操作原则
- 不操作用户已有 tab,在自建后台 tab 中操作
- 完成后
/close自建 tab - 先了解页面结构,再决定下一步
- GUI 交互产生的链接是可靠的(携带完整上下文)
并行调研:子 Agent 分治
多个独立目标时,分发子 Agent 并行执行。
| 适合分治 | 不适合分治 |
| 目标独立、互不依赖 | 目标有依赖关系 |
| 每个子任务量足够大 | 简单查询 |
| 需要 CDP 或长时间运行 | 轻量搜索 |
子 Agent prompt 要求:描述目标(获取、调研、了解),不暗示具体手段。
信息核实
核实的目标是一手来源,非二手报道。
| 信息类型 | 一手来源 |
| 政策/法规 | 发布机构官网 |
| 企业公告/财报 | 交易所/企业官网 |
| 学术声明 | 原始论文/机构官网 |
| 行业数据 | 数据发布方 |
找不到官方原文时,标明来源局限性。
站点经验
按域名存储在 skills/web-access/references/site-patterns/ 下。
- 确定目标网站后,检查是否有匹配经验文件
- 经验是"可能有效的提示",非保证
- 操作成功后,发现新模式主动写入经验文件
References 索引
| 文件 | 何时加载 |
skills/web-access/references/cdp-api.md | 需要 CDP API 详细参考时 |
skills/web-access/references/site-patterns/{domain}.md | 确定目标网站后 |
skills/web-access/references/migration-2.5.3.md | 遇到旧写法迁移提示时 |
行为准则
- 搜索不是终点:搜到 URL 后必须获取内容或下载文件
- 能自动化的不推给用户:只有登录态/验证码/付费墙才标记"需手动"
- 下载后验证:file + wc -c
- 失败时切换通道:不在同一方式上反复重试
- 记录来源:所有获取的数据记录 URL 和时间
- 不静默降级:降级时告知用户原因
- 站点经验积累:成功操作后更新经验文件
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

