Pular para o conteúdo
TakoAPI
Voltar às skills

grago

por solsuk

Browser & Automation
282Downloads informados pelo clawskills.sh
0Visualizações
0 curtidaEntre para curtir esta skill
—Ainda sem avaliações
Sua avaliação

Resumo

**Delegate research and data-fetch tasks to a free local LLM.

Descrição

O que esta skill faz

Grago bridges OpenClaw agents and local LLMs like Ollama or llama.cpp that can't use tools natively. It runs shell scripts to pull live data from the web, APIs, or local files, then pipes the results into a local model with a focused prompt. The cloud model handles reasoning; the local machine handles the fetch-and-analyze work.

It offloads token-heavy fetch-and-analyze work to a local LLM, cutting API costs while still giving the agent access to live external data.

Quando usá-la

  • Fetching and summarizing web pages without spending cloud API tokens
  • Analyzing API responses locally to keep data off third-party servers
  • Monitoring RSS feeds and extracting key updates with a local model
  • Running multi-source research jobs on local hardware overnight
  • Piping log files through a local LLM to detect patterns or anomalies

Exemplo de fluxo de trabalho

ENTRADA

User asks: Fetch this API endpoint and identify any trends in the results

AGENTE
  1. 1

    Grago pipes the shell command curl -s https://api.example.com/data to fetch live data

  2. 2

    Transforms the response with jq .results to extract relevant fields

  3. 3

    Sends the transformed output to the local Ollama model with the prompt 'Identify trends and flag outliers'

  4. 4

    Returns the model's structured analysis back to the OpenClaw agent

SAÍDA

A markdown summary of trends and flagged outliers from the API data, generated locally without cloud tokens

Requisitos

  • Ollama installed and running locally
  • At least one model pulled in Ollama (e.g., gemma2, mistral, llama3)

Instalação

clawhub install solsuk/grago

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

Links