smart-web-fetch-safe安全版智能网页内容获取技能。本地解析 + 可选远程清洗,隐私优先,Token 优化。
Install via ClawdBot CLI:
clawdbot install vimself/smart-web-fetch-safeGrade Fair — based on market validation, documentation quality, package completeness, maintenance status, and authenticity signals.
Calls external URL not in known-safe list
https://example.com/articleUses known external API (expected, informational)
r.jina.aiAudited Apr 18, 2026 · audit v1.0
Generated Oct 7, 2026
A journalist or analyst configures the skill in local mode with a strict ALLOWED_DOMAINS whitelist to scrape and summarize public articles. Because all parsing happens in-process, sensitive queries never leave the machine. Useful when researching topics that might reveal a client's interests or an ongoing investigation.
A developer builds a daily Slack bot that fetches 20+ news URLs and condenses them into a single digest. Running with --remote --max-chars 3000 keeps the LLM context small and cheap, since Jina's reader strips ads, navbars, and footers before the model ever sees the text.
An e-commerce team monitors competitor product pages on an approved domain list and extracts key specs and price blocks. Local parsing with beautifulsoup4 gives precise CSS-selector control when the remote Jina output is too lossy, while --max-chars prevents accidental full-page dumps into the database.
A grad student pipes arXiv and Wikipedia URLs through the tool to build a reading summary corpus. Local mode respects publisher terms, and the domain whitelist prevents the agent from wandering into unrelated or paywalled sites when given a loose prompt.
An internal SOC scripted the skill to fetch suspicious URLs submitted by employees, but forces local mode only so link contents are never relayed to a third party. The ALLOWED_DOMAINS allowlist blocks the agent from being prompt-injected into fetching attacker-controlled domains outside policy.
Open-source the local-parsing core and sell tiers that unlock the remote Jina reader, higher MAX_CHARS ceilings, and per-team ALLOWED_DOMAINS policies. The free tier drives adoption among indie devs while paid seats come from teams that need token savings at scale.
Host a managed version of the remote-clean mode behind a metered API, charging per 1,000 fetched pages or per million tokens of cleaned text. Customers integrate it directly into their own agent pipelines instead of running Python locally, trading privacy for convenience and clean output.
Package the local-mode variant plus audit logs, DLP hooks, and configurable domain policies as a compliance product for regulated industries. The selling point is that no URL or content ever leaves the corporate network, satisfying data-residency requirements that disqualify cloud scrapers.
💬 Integration Tip
Start with local mode and an explicit ALLOWED_DOMAINS whitelist, since the script's default '*' is unsafe for untrusted agent-driven input; only enable --remote when token cost outweighs the third-party privacy tradeoff. Wrap calls with a hard --max-chars cap and parse the --json output programmatically to keep downstream context bounded.
Scored Oct 7, 2026
AI-optimized web search via Tavily API. Returns concise, relevant results for AI agents.
Search the web using Baidu AI Search Engine (BDSE). Use for live information, documentation, or research topics.
Automatically fetch YouTube video transcripts, generate structured summaries, and send full transcripts to messaging platforms. Detects YouTube URLs and provides metadata, key insights, and downloadable transcripts.
Search the web with AI-powered answers via Perplexity API. Returns grounded responses with citations. Supports batch queries.
Fetches top stories from TechMeme, summarizes linked articles, and highlights social media reactions. Use when user wants tech news or says /technews.
黄历查询工具。按公历日期查询今日黄历、老黄历、每日宜忌、吉神凶煞、神煞、冲煞、吉日凶日。适合查询今天黄历怎么样、今天宜忌、明天适不适合搬家、结婚开业开工选哪天、最近哪些是吉日。支持单日黄历、批量区间查询、多日筛选吉日。