← 返回 Skills 市场
neverchenx

Agent Reach

作者 Never · GitHub ↗ · v1.1.0 · MIT-0
cross-platform ⚠ suspicious
413
总下载
0
收藏
1
当前安装
1
版本数
在 OpenClaw 中安装
/install agent-reach-en
功能描述
Give your AI agent eyes to see the entire internet. Install and configure upstream tools for Twitter/X, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu, Douyi...
使用说明 (SKILL.md)

Agent Reach

Install and configure upstream tools for 13+ platforms. After setup, call them directly — no wrapper layer.

Workspace Rules

Never create files, clone repos, or write output in the agent workspace. Use these directories instead:

Purpose Directory
Temporary output (subtitles, downloads) /tmp/
Upstream tool repos ~/.agent-reach/tools/
Config & tokens ~/.agent-reach/

Violating this will pollute the user's workspace and degrade their agent experience over time.

Setup

pip install https://github.com/Panniantong/agent-reach/archive/main.zip
agent-reach install --env=auto
agent-reach doctor

install auto-detects your environment and installs core dependencies (Node.js, mcporter, xreach CLI, gh CLI, yt-dlp, feedparser). Run doctor to see what's active.

Management

agent-reach doctor        # channel status overview
agent-reach watch         # quick health + update check
agent-reach check-update  # check for new versions

Configure channels

agent-reach configure twitter-cookies "auth_token=xxx; ct0=yyy"
agent-reach configure proxy http://user:pass@ip:port
agent-reach configure --from-browser chrome    # auto-extract cookies from local browser

Configuring a channel

When a user asks to configure/enable any channel:

  1. Run agent-reach doctor
  2. Find the channel — it shows status (✅/⚠️/⬜) and what to do next
  3. Execute what you can automatically (install packages, start services)
  4. For human-required steps (paste cookies), tell the user what to do
  5. Run agent-reach doctor again to verify

Do NOT memorize per-channel steps. Always rely on doctor output.

Cookie Import (Universal for all login-required platforms)

Important: Platforms using cookie login have account ban risks. Remind users to use a dedicated secondary account.

For all platforms requiring cookies (Twitter, XiaoHongShu, etc.), prefer Cookie-Editor import:

  1. User logs into the platform in their browser
  2. Install Cookie-Editor Chrome extension
  3. Click extension → Export → Header String
  4. Send the exported string to the Agent

Local computer users can also use agent-reach configure --from-browser chrome for automatic extraction.

QR code login is a backup option (only for local environments with browser), Cookie-Editor is simpler and more reliable.

Other human actions

  • Proxy: Reddit/Bilibili/XiaoHongShu may block server IPs — suggest a residential proxy if on a server

Using Upstream Tools Directly

After agent-reach install, call the upstream tools directly.

Note: agent-reach is an installer and config tool — it does NOT have read, search, or content-fetching commands. Use the upstream tools below instead.

Twitter/X (xreach CLI)

# Search tweets
xreach search "query" --json -n 10

# Read a specific tweet
xreach tweet https://x.com/user/status/123 --json

# Read a user's timeline
xreach tweets @username --json -n 20

YouTube (yt-dlp)

yt-dlp requires JS runtime for YouTube downloads. agent-reach install auto-configures Node.js as runtime. If you see "Sign in to confirm you're not a bot", the IP is blocked by YouTube anti-crawler. Switch proxy or add cookies.

# Get video metadata
yt-dlp --dump-json "https://www.youtube.com/watch?v=xxx"

# Download subtitles only
yt-dlp --write-sub --write-auto-sub --sub-lang "zh-Hans,zh,en" --skip-download -o "/tmp/%(id)s" "URL"
# Then read the .vtt file

# Search (yt-dlp ytsearch)
yt-dlp --dump-json "ytsearch5:query"

# If "no JS runtime" warning: ensure Node.js is installed, then run:
#   mkdir -p ~/.config/yt-dlp && echo "--js-runtimes node" >> ~/.config/yt-dlp/config

Bilibili (yt-dlp)

Server IPs may be blocked by Bilibili (412 error). Recommend using proxy or add --cookies-from-browser chrome.

# Get video metadata
yt-dlp --dump-json "https://www.bilibili.com/video/BVxxx"

# Download subtitles
yt-dlp --write-sub --write-auto-sub --sub-lang "zh-Hans,zh,en" --convert-subs vtt --skip-download -o "/tmp/%(id)s" "URL"

# If blocked (412 / login required):
yt-dlp --cookies-from-browser chrome --dump-json "URL"

Reddit (JSON API)

# Read a subreddit
curl -s "https://www.reddit.com/r/python/hot.json?limit=10" -H "User-Agent: agent-reach/1.0"

# Read a post with comments
curl -s "https://www.reddit.com/r/python/comments/POST_ID.json" -H "User-Agent: agent-reach/1.0"

# Search
curl -s "https://www.reddit.com/search.json?q=query&limit=10" -H "User-Agent: agent-reach/1.0"

Note: On servers, Reddit may block your IP. Use proxy or search via Exa instead.

XiaoHongShu (mcporter + xiaohongshu-mcp)

Requires login. Import cookies via Cookie-Editor or QR code login.

# Search posts
mcporter call 'xiaohongshu.search_feeds(keyword: "query")'

# Get post details (with comments)
mcporter call 'xiaohongshu.get_feed_detail(feed_id: "xxx", xsec_token: "yyy")'

# Get all comments
mcporter call 'xiaohongshu.get_feed_detail(feed_id: "xxx", xsec_token: "yyy", load_all_comments: true)'

# Publish image post
mcporter call 'xiaohongshu.publish_content(title: "Title", content: "Content", images: ["/path/to/img.jpg"], tags: ["food"])'

# Publish video post
mcporter call 'xiaohongshu.publish_with_video(title: "Title", content: "Content", video: "/path/to/video.mp4", tags: ["vlog"])'

Other features (like, collect, comment, user profile, etc.): npx mcporter list xiaohongshu

Douyin (mcporter + douyin-mcp-server)

# Parse Douyin video info (share link -> title, author, watermark-free video URL, etc.)
mcporter call 'douyin.parse_douyin_video_info(share_link: "https://v.douyin.com/xxx/")'

# Get watermark-free video download link
mcporter call 'douyin.get_douyin_download_link(share_link: "https://v.douyin.com/xxx/")'

# AI extract video voice transcript (requires SiliconFlow API Key)
mcporter call 'douyin.extract_douyin_text(share_link: "https://v.douyin.com/xxx/")'

No login required to parse videos. Supports both Douyin share links and direct links.

GitHub (gh CLI)

# Search repos
gh search repos "query" --sort stars --limit 10

# View a repo
gh repo view owner/repo

# Search code
gh search code "query" --language python

# List issues
gh issue list -R owner/repo --state open

# View a specific issue/PR
gh issue view 123 -R owner/repo

Web — Any URL (Jina Reader)

# Read any webpage as markdown
curl -s "https://r.jina.ai/URL" -H "Accept: text/markdown"

# Search the web
curl -s "https://s.jina.ai/query" -H "Accept: text/markdown"

Exa Search (mcporter + exa MCP)

# Web search
mcporter call 'exa.web_search_exa(query: "query", numResults: 5)'

# Code search (GitHub, StackOverflow, docs)
mcporter call 'exa.get_code_context_exa(query: "how to parse JSON in Python", tokensNum: 3000)'

# Company research
mcporter call 'exa.company_research_exa(companyName: "OpenAI")'

LinkedIn (mcporter + linkedin-scraper-mcp)

# View a profile
mcporter call 'linkedin.get_person_profile(linkedin_url: "https://linkedin.com/in/username")'

# Search people
mcporter call 'linkedin.search_people(keyword: "AI engineer", limit: 10)'

# View company
mcporter call 'linkedin.get_company_profile(linkedin_url: "https://linkedin.com/company/xxx")'

Fallback: curl -s "https://r.jina.ai/https://linkedin.com/in/username"

Boss Zhipin (mcporter + mcp-bosszp)

# Browse recommended jobs
mcporter call 'bosszhipin.get_recommend_jobs_tool(page: 1)'

# Search jobs
mcporter call 'bosszhipin.search_jobs_tool(keyword: "Python", city: "Beijing", page: 1)'

# View job details
mcporter call 'bosszhipin.get_job_detail_tool(job_url: "https://www.zhipin.com/job_detail/xxx")'

Fallback: curl -s "https://r.jina.ai/https://www.zhipin.com/job_detail/xxx"

WeChat Official Accounts (wechat-article-for-ai + miku_ai)

Search (miku_ai — Sogou WeChat search):

# Search WeChat articles by keyword
python3 -c "
import asyncio
from miku_ai import get_wexin_article

async def search():
    articles = await get_wexin_article('AI Agent', 5)
    for a in articles:
        print(f'{a[\"title\"]} | {a[\"source\"]} | {a[\"date\"]}')
        print(f'  {a[\"url\"]}')

asyncio.run(search())
"

Read (Camoufox — stealth Firefox, bypasses WeChat anti-bot):

# Read a WeChat article (returns Markdown with images)
cd ~/.agent-reach/tools/wechat-article-for-ai && python3 main.py "https://mp.weixin.qq.com/s/ARTICLE_ID"

# Run as MCP server (for AI agent integration)
python3 mcp_server.py

Typical agent workflow: search → get URLs → immediately read full content.

Note: WeChat articles require a real browser to render. Jina Reader and curl cannot read them.

RSS (feedparser)

python3 -c "
import feedparser
d = feedparser.parse('https://example.com/feed')
for e in d.entries[:5]:
    print(f'{e.title} — {e.link}')
"

Troubleshooting

Twitter "fetch failed"

xreach CLI uses Node.js undici, which doesn't respect HTTP_PROXY. Solutions:

  1. Ensure undici is installed: npm install -g undici
  2. Configure proxy: agent-reach configure proxy http://user:pass@ip:port
  3. If still failing, use transparent proxy (Clash TUN, Proxifier)

Channel broken?

Run agent-reach doctor — it shows what's wrong and how to fix it.

Changelog

Version Date Changes
v1.1.0 2025-03-15 Translated to English
v1.0.0 2025-03-05 Initial release
安全使用建议
Proceed with caution. This skill legitimately needs tooling to read many sites, but it also asks you to provide full browser cookie/header strings (which can log you in as that user). Never paste cookies from a primary/personal account — use a dedicated secondary account. If possible, prefer OAuth tokens or API keys scoped to the minimum permissions instead of session cookies. Before running pip install from the GitHub zip, inspect the repository source and prefer installing inside an isolated environment (virtualenv, container). If using automatic cookie extraction, only do so on a personal/local machine you control and understand where ~/.agent-reach will store credentials; consider encrypting or restricting filesystem permissions. Finally, ask the skill author to (1) declare required credentials in the registry metadata, (2) explain storage/encryption of tokens, and (3) provide a vetted release (tagged GitHub release or published package) rather than an archive from main.
功能分析
Type: OpenClaw Skill Name: agent-reach-en Version: 1.1.0 The skill bundle automates the installation of numerous third-party tools and the collection of sensitive session cookies for 13+ platforms, including a command to extract cookies directly from the local Chrome browser (`agent-reach configure --from-browser chrome`). While these actions support the stated goal of platform integration, the automated handling of credentials and the execution of remote scripts from GitHub (via `pip install` of a ZIP) represent high-risk behaviors. No explicit evidence of exfiltration was found in the instructions, but the broad access to user accounts makes it suspicious.
能力评估
Purpose & Capability
The skill's name/description (provide access to many platforms) aligns with the actions described (install platform CLIs, configure channels, use yt-dlp, xreach, mcporter, etc.). Requesting Node.js, yt-dlp, gh, and other tooling is proportionate to the aggregator role. However, the skill does not declare any credentials in the registry metadata while its runtime instructions clearly require handing over sensitive session cookies and tokens — a mismatch worth flagging.
Instruction Scope
SKILL.md explicitly instructs the agent/user to export full cookie 'Header String' from a browser extension or use automatic cookie extraction (--from-browser chrome) and then send those cookie strings to the Agent. That is effectively giving the agent account authentication material for many services. The instructions also encourage running commands that will write configs and tokens under ~/.agent-reach and using proxies with potentially embedded credentials. Asking users to paste or allowing automatic extraction of browser cookies is broad and sensitive and is not limited/scoped or constrained in the skill metadata.
Install Mechanism
There is no formal install spec in the registry, but SKILL.md tells users to pip install a GitHub archive URL (github.com/Panniantong/agent-reach/archive/main.zip) which will execute arbitrary Python package installation on the user's machine. Using a direct GitHub zip is common but higher-risk than installing a vetted package from an official registry; the skill also instructs installing multiple external tools (Node, yt-dlp, mcporter, xreach) which is expected for its purpose but increases the attack surface.
Credentials
The registry lists no required credentials, yet the instructions rely heavily on users providing session cookies and tokens (via Cookie-Editor export or automatic extraction). Sensitive data transfer is surfaced only in SKILL.md text, not as declared required env vars or a primary credential. This disconnect makes it hard for users to audit what secrets the skill will handle and how they'll be stored or protected.
Persistence & Privilege
The skill does not request 'always: true' and is user-invocable (normal). It instructs storing tool repos and config/tokens under ~/.agent-reach, which is consistent with its installer role but means long-lived credentials will be written to the user's home directory. The skill does not explain encryption, permissions, or secure storage of those secrets.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install agent-reach-en
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /agent-reach-en 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.1.0
Translated to English
元数据
Slug agent-reach-en
版本 1.1.0
许可证 MIT-0
累计安装 1
当前安装数 1
历史版本数 1
常见问题

Agent Reach 是什么?

Give your AI agent eyes to see the entire internet. Install and configure upstream tools for Twitter/X, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu, Douyi... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 413 次。

如何安装 Agent Reach?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install agent-reach-en」即可一键安装,无需额外配置。

Agent Reach 是免费的吗?

是的,Agent Reach 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Agent Reach 支持哪些平台?

Agent Reach 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Agent Reach?

由 Never(@neverchenx)开发并维护,当前版本 v1.1.0。

💬 留言讨论