Pular para o conteúdo
TakoAPI
Voltar às skills

friday-router

por runeweaverstudios

Coding Agents & IDEs
540Downloads informados pelo clawskills.sh
0Visualizações
0 curtidaEntre para curtir esta skill
—Ainda sem avaliações
Sua avaliação

Resumo

Your AI's Smart Traffic Director—precisely matching OpenClaw tasks to the perfect LLM.

Descrição

O que esta skill faz

IntentRouter classifies task descriptions by type and routes them to a specific LLM via OpenRouter. It runs as an orchestrator on Gemini 2.5 Flash, delegating code tasks to MiniMax 2.5, creative tasks to Kimi k2.5, and research tasks to Grok Fast. Router output contains only task parameters and model selection, no auth tokens or gateway credentials.

Instead of manually selecting a model per task in OpenClaw, the router classifies intent and spawns the right sub-agent automatically using a single OpenRouter auth profile.

Quando usá-la

  • Routing a bug fix request to a code-specialized model without manual selection
  • Sending a creative writing prompt to Kimi k2.5 automatically
  • Delegating research queries to Grok Fast for web-search tasks
  • Handling mixed daily OpenClaw workflows without picking a model each time
  • Running image analysis tasks through a GPT-4o sub-agent

Exemplo de fluxo de trabalho

ENTRADA

User asks: Write a poem about the ocean

AGENTE
  1. 1

    Run router.py spawn --json 'Write a poem about the ocean' to classify the task

  2. 2

    Receive JSON output specifying model openrouter/moonshotai/kimi-k2.5 and sessionTarget

  3. 3

    Call sessions_spawn with the exact task and model values from the router output

  4. 4

    Strip any internal orchestration metadata from the sub-agent response

  5. 5

    Forward the poem to the user labeled 'Using: Kimi k2.5'

SAÍDA

A poem generated by the Kimi k2.5 sub-agent, presented cleanly without session metadata

Requisitos

  • OpenRouter API key configured in OpenClaw auth profile

Instalação

clawhub install runeweaverstudios/friday-router

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

Links