登录
MattHuanglol

model-priority

by @MattHuanglol in AI & LLM
0
0
# Install this skill:
npx skills add MattHuanglol/openclaw_skill --skill "model-priority"

Install specific skill from multi-skill repository

# Description

Manage model selection priority and usage guidance. Use when the user asks which model to use, when switching models for specific tasks, or when managing usage limits and fallbacks. Includes a full list of available models and their typical use cases.

# SKILL.md

name: model-priority
description: Manage model selection priority and usage guidance. Use when the user asks which model to use, when switching models for specific tasks, or when managing usage limits and fallbacks. Includes a full list of available models and their typical use cases.

Model Priority & Selection Guide

This skill standardizes how to select the best AI model for different tasks within Matt’s OpenClaw environment.

Available Models List

Provider/ID Display Name Best For
google-antigravity/gemini-3-pro-high Gemini 3 Pro (High) Complex reasoning, creative writing, high-accuracy tasks.
google-antigravity/gemini-3-pro-low Gemini 3 Pro (Low) Faster pro-level reasoning with lower token cost/latency.
google-antigravity/gemini-3-flash Gemini 3 Flash High speed, summarizing, routine data extraction.
google-antigravity/claude-sonnet-4-5 Claude 3.5 Sonnet Programming, complex instructions, logical consistency.
google-antigravity/claude-sonnet-4-5-thinking Claude 3.5 Sonnet (Think) Deep technical analysis, debugging, planning.
google-antigravity/claude-opus-4-5-thinking Claude 3.5 Opus (Think) Maximum intelligence for extremely complex problems.
google-antigravity/gpt-oss-120b-medium GPT-OSS 120B Open-source large model alternative for general tasks.
google-gemini-cli/gemini-3-pro-preview Gemini Pro (CLI) High-priority for one-shot Q&A and coding tasks via direct OAuth.
google-gemini-cli/gemini-2.5-pro Gemini 2.5 Pro (CLI) Solid alternative for general tasks via direct CLI OAuth.
google-gemini-cli/gemini-2.5-flash Gemini 2.5 Flash (CLI) High-speed, cost-effective 2.5 series model via CLI.
google-gemini-cli/gemini-2.5-flash-lite Gemini 2.5 Flash Lite (CLI) Ultra-fast, lightweight model for quick tasks via CLI.
google-gemini-cli/gemini-3-flash-preview Gemini Flash (CLI) Fast CLI-based tasks and status checks.
openai-codex/gpt-5.2 Codex (ChatGPT) General coding and multi-turn conversation (Team plan).

Selection Logic (Priority)

  1. 日常/一般任務

    • 主要：openai-codex/gpt-5.2 (ChatGPT)
    • 備援：google-antigravity/gemini-3-pro-low

  2. 程式開發與重構

    • 主要：google-antigravity/claude-sonnet-4-5
    • 備援：openai-codex/gpt-5.2

  3. 背景任務 (Cron/Heartbeat)

    • 主要：google-antigravity/gemini-3-flash
    • 備援：google-gemini-cli/gemini-3-flash-preview

  4. 複雜規劃與架構設計

    • 主要：google-antigravity/gemini-3-pro-high 或 claude-opus-4-5-thinking
    • 備援：google-antigravity/claude-sonnet-4-5-thinking 或 google-gemini-cli/gemini-3-pro-preview

How to Switch Models

  • Session Switch: /model <provider/id>
  • Background Switch (Cron): Update the payload.model field in the cron job configuration.

Usage Limit Management

If a model reports a usage limit error (e.g., "ChatGPT usage limit hit"):
1. Check 📊 session_status to see remaining credits for other providers.
2. Manually switch the session or background job to a provider with 100% capacity (usually google-antigravity).

# Related Skills

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

⚡ Amp 🚀 Antigravity 🤖 Claude Code 🦀 Clawdbot 📝 Codex ▶️ Cursor 🤖 Droid 💎 Gemini CLI 🐙 GitHub Copilot 🪿 Goose 📊 Kilo Code 🔧 Kiro CLI 💻 OpenCode 🦘 Roo Code 🌲 Trae 🏄 Windsurf

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.