Url Reader

智能读取任意URL内容，支持微信公众号、小红书、今日头条、抖音、淘宝、天猫、京东、百度等中国主流平台，自动识别平台类型并提取核心内容。自动保存内容为Markdown，下载图片到本地。

What to consider before installing: - The skill does what it claims (scrape many Chinese platforms and save Markdown/images) but is sloppy about metadata: it does require an optional FIRECRAWL_API_KEY (documented in SKILL.md) and Playwright, even though the registry lists no env vars. Treat the Firecrawl API key as sensitive. - The skill will write files to disk: it uses a hard-coded default save directory (/Users/ys/laoyang知识库/nickys/素材). Edit the DEFAULT_OUTPUT_DIR in scripts/save_content.py and url_reader.py before use to point to a directory you control, or run the scripts from a confined/sandboxed environment. - For sites that require login (WeChat, Taobao, etc.), the skill uses Playwright to launch a browser and will save browser storage_state (cookies/session tokens) to data/wechat_auth.json inside the skill directory. Those files contain authentication data — review and store them securely, or avoid using the Playwright login flow if you don't want to persist credentials. - Installing Playwright will download Chromium binaries (playwright install chromium). Only proceed if you are comfortable with that and run the install in a controlled environment (virtualenv, container, or VM). - The skill contacts external services: Firecrawl (requires API key, paid tiers) and r.jina.ai (free). If you do not trust Firecrawl, do not set its API key; the skill will fallback to Jina/Playwright but with degraded behavior. - Because the repository owner and homepage are unknown, exercise extra caution. Recommended steps before installing: (1) edit DEFAULT_OUTPUT_DIR to a safe location, (2) confirm or remove automatic saving of wechat_auth.json if you do not want local credential persistence, (3) only provide FIRECRAWL_API_KEY if you trust the service and understand billing, and (4) run the tool inside a sandbox (container or VM) until you are comfortable with its behavior. If the author updated the package metadata to declare FIRECRAWL_API_KEY as an optional required env var, and replaced the hard-coded output path with a configurable default or documented prompt, my assessment would move toward 'benign'.

Type: OpenClaw Skill Name: url-reader Version: 0.1.1 The skill is designed to read and save content from arbitrary URLs, which inherently involves network requests and local file system writes. It is classified as 'suspicious' due to several vulnerabilities rather than clear malicious intent. Key indicators include a hardcoded output directory (`/Users/ys/laoyang知识库/nickys/素材`) in `skill.md`, `scripts/save_content.py`, and `scripts/url_reader.py`, which could lead to unintended file writes. Furthermore, `scripts/save_content.py` and `scripts/url_reader.py` download images from arbitrary URLs and save them to disk without robust content validation, posing a risk of downloading malicious files. The use of Playwright's `page.evaluate()` in `scripts/url_reader.py`, `scripts/wechat_reader.py`, and `scripts/wechat_reader_v2.py` to execute JavaScript in a browser context could also be vulnerable to client-side injection if a malicious URL is processed. While these are significant risks, there is no evidence of intentional data exfiltration, backdoor installation, or harmful prompt injection.