Pular para o conteúdo
TakoAPI
Voltar às skills

ezcto-smart-web-reader

por takahashigy

Web & Frontend Development
447Downloads informados pelo clawskills.sh
2Estrelas
0Visualizações
0 curtidaEntre para curtir esta skill
—Ainda sem avaliações
Sua avaliação

Resumo

Agent web access acceleration layer — reads any URL as structured JSON.

Descrição

O que esta skill faz

A web access layer for OpenClaw agents that reads any URL and returns structured JSON containing page identity, content sections, media metadata, and actionable links. Uses a cache-first strategy: if the EZCTO community library has the page, the result returns in ~1 second at zero token cost. On a cache miss, it fetches the HTML, detects the site type, and parses it with an LLM.

Cache hits cost zero tokens and complete in under two seconds, versus 500-2000 tokens per request with raw web_fetch or screenshot-based approaches.

Quando usá-la

  • Checking what a shared URL contains without leaving the chat
  • Extracting product details and prices from an e-commerce page
  • Getting structured data from a crypto token or DeFi project site
  • Looking up a restaurant's hours, menu, and reservation links
  • Researching a company's website before drafting an outreach message

Exemplo de fluxo de trabalho

ENTRADA

User asks: What's on this page? https://example.com/product

AGENTE
  1. 1

    Compute URL hash and check EZCTO community cache via api.ezcto.fun

  2. 2

    On cache miss, fetch HTML with curl and compute SHA256 hash

  3. 3

    Run pattern matching to detect site type (crypto, ecommerce, restaurant, or general)

  4. 4

    Strip scripts, styles, and comments from HTML, then parse with LLM into structured JSON

  5. 5

    Store result locally in ~/.ezcto/cache/ and contribute to EZCTO asset library

SAÍDA

Structured JSON with page metadata, content sections, detected entities, media descriptions, action links, and suggested next steps for the agent

Instalação

clawhub install takahashigy/ezcto-smart-web-reader

Execute o comando da CLI ou cole o prompt no Claude Code, Cursor, Codex ou no agente de código que você usa.

Links