Ollama Model Agents

Use this skill when Kevin asks for Gemini CLI, Ollama cloud/local models, DeepSeek, Qwen/QWIN, GLM, or Claude Code routed through Ollama. Keep Codex responsible for final synthesis and validation.

Backends

• Gemini CLI: current research, long-context synthesis, independent advisory review.
• Direct Ollama: DeepSeek, Qwen, GLM, Gemma, Llama, local survival-mode opinions.
• Claude Code via Ollama: Claude Code harness using an Ollama model through Anthropic-compatible API.

Required Checks

1. Run ollama list to confirm installed or cloud-available models.
2. Prefer a dry run first when building commands for Linda.
3. Do not use API keys unless Kevin explicitly approves a specific emergency task. Subscription-authenticated CLIs, Ollama cloud entitlement, and local Ollama models are the default.
4. Strip paid provider API-key environment variables before delegated model calls: ANTHROPIC_API_KEY, OPENAI_API_KEY, GEMINI_API_KEY, GOOGLE_API_KEY, OPENROUTER_API_KEY, XAI_API_KEY, DEEPSEEK_API_KEY, QWEN_API_KEY, DASHSCOPE_API_KEY, GLM_API_KEY, and ZHIPUAI_API_KEY.
5. For Claude Code through Ollama, set: