agentleFS
Sign inStart free
OpenMinisOpenMinis/MinisSkills/web-search/SKILL.md

web-search

Skill584 starsChanged 59 days agoMITView on GitHub

General web search skill that uses browser automation with multiple search engines and selects the priority chain based on search intent. Supports: Perplexity, Metaso AI, Google, Bing, Brave, DuckDuckGo, Baidu, Sogou, Tavily. Trigger when the user mentions "web search," "search for it," "search this for me," "look it up online," "search engine," or needs real-time information from the internet.

We split this skill into the rules it gives an agent

10 rules. 0 are common, 10 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 63 edited versions we read keep these.

  • 10

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use browser automation to search the web through multiple engines, then fall back automatically when one engine is blocked, logged out, or …

    line 12
  2. Distinctive:

    search engine results instead of model memory

    line 18
  3. Distinctive:

    only input box is visible with no actual results

    line 61
  4. Distinctive:

    Treat a search as successful when at least two of these are true:

    line 65
  5. Distinctive:

    If blocked or poor, switch to the next engine.

    line 78
  6. Distinctive:

    Prefer Perplexity for best answer quality when login is valid.

    line 88
  7. Distinctive:

    Prefer Metaso AI for Chinese structured answers.

    line 89
  8. Distinctive:

    Prefer Google/Bing for finding original pages and official websites.

    line 90
  9. Distinctive:

    Use Brave/DuckDuckGo as privacy-oriented backups.

    line 91
  10. Distinctive:

    Do not rely on a single engine.

    line 92

“Common” is measured: kept by at least 6 in 10 of the 63 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

web-search

Use browser automation to search the web through multiple engines, then fall back automatically when one engine is blocked, logged out, or returns poor results.

When to use

Activate this skill when the user wants:

  • real-time web search
  • search engine results instead of model memory
  • Chinese web search or structured AI answers
  • engine-specific search (Google / Bing / Brave / Perplexity / Metaso AI)
  • fallback behavior across multiple search engines

Search engines

EngineURL patternStrengthNotes
Perplexityhttps://www.perplexity.ai/search?q={query}Highest answer qualityNeeds login once
Metaso AIhttps://metaso.cn/Strong Chinese resultsHome page interaction may be needed
Googlehttps://www.google.com/search?q={query}Best general web coverageStable fallback
Binghttps://www.bing.com/search?q={query}Good Chinese + AI summaryStable fallback
Bravehttps://search.brave.com/search?q={query}Privacy-friendlyGood backup
DuckDuckGohttps://html.duckduckgo.com/html/?q={query}Lightweight privacy searchMay show anomaly pages
Baiduhttps://www.baidu.com/s?wd={query}Strong Chinese indexUse for Chinese web results
Sogouhttps://www.sogou.com/web?query={query}Chinese / WeChat-heavy resultsSecondary Chinese backup
TavilyAPI-basedAgent-orientedRequires API key

Intent-based priority chains

Choose the search chain by user intent.

Deep research / comparison / trend analysis

Perplexity -> Metaso AI -> Bing -> Google -> Brave

Chinese search / Chinese summary / local context

Metaso AI -> Perplexity -> Baidu -> Bing -> Google

General web search / official site / original page lookup

Google -> Bing -> Brave -> DuckDuckGo

Privacy-first search

Brave -> DuckDuckGo -> Bing

Fallback rules

Immediately switch to the next engine when any of these happens:

  • login expired
  • captcha / verification / blocked / robot / abnormal traffic
  • empty result page
  • redirected back to the home page
  • page structure is broken and no useful text can be extracted
  • only input box is visible with no actual results

Success criteria

Treat a search as successful when at least two of these are true:

  • the page title or URL clearly indicates a result page
  • extracted text is non-empty and contains meaningful results
  • there are structured answers, result lists, citations, or related questions
  • the content is clearly not just the search homepage

Execution flow

  1. Infer search intent from the user request.
  2. Select the matching priority chain.
  3. Try the first engine.
  4. Extract text with browser tools.
  5. Check success criteria.
  6. If blocked or poor, switch to the next engine.
  7. Return the first good result, and optionally add a second source for comparison.

Scripts

  • scripts/browser_search.py - generate engine priority plans and fallback chains
  • evals/evals.json - prompt coverage for the main routing scenarios

Notes

  • Prefer Perplexity for best answer quality when login is valid.
  • Prefer Metaso AI for Chinese structured answers.
  • Prefer Google/Bing for finding original pages and official websites.
  • Use Brave/DuckDuckGo as privacy-oriented backups.
  • Do not rely on a single engine; fallback is part of the skill design.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

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.

Check your setup, free