ollama-cli-ask

通过本地的‘ask’ CLI，与Ollama模型进行交互。支持一次性查询、交互式对话、上下文传递，以及JSON格式的输出。

SKILL.md

--- frontmatter

name: ollama-cli-ask
description: Use the local 'ask' CLI to interact with Ollama models. Supports one-shot queries, interactive chat, piping context, and JSON output.
purpose: Fast, local LLM inference without cloud costs or latency.
usage: |
  # One-shot
  ask "Question"
  
  # Pipe context
  cat file.txt | ask "Summarize this"
  
  # Interactive
  ask
  
  # Specific model
  ask -m qwen3-coder-next "Code me a snake game"
  
  # Reasoning/Thinking mode
  ask -t "Think step by step"
parameters:
  - prompt: The prompt to send (optional, enters interactive mode if empty)
  - model: (Optional) Model to use (default: qwen3-coder-next)
  - system: (Optional) System prompt
  - think: (Optional) Enable thinking/reasoning mode
  - json: (Optional) Force JSON output
dependencies:
  - ollama (running)
  - ~/bin/ask (installed CLI)
created: 2026-02-04

Ollama CLI (ask) Skill 🦑

This skill leverages the ask CLI tool to interact with local Ollama models. It's faster and more flexible than raw curl requests.

Features

•Context-aware: Remembers conversation history in interactive mode.
•Piping: Can ingest file content or command output via stdin.
•Thinking: Supports <think> tag parsing for reasoning models (DeepSeek, etc).
•JSON: reliable JSON output for tool use.

Installation (if missing)

bash

cp ~/clawd/tools/ollama-cli/ask.sh ~/bin/ask
chmod +x ~/bin/ask

Recommended Models

•Coding: qwen3-coder-next (Smartest local coder)
•General: llama3.1:8b (Fast, good generalist)
•Reasoning: deepseek-r1 (if available, use -t flag)

Examples

Summarize a file:

bash

cat MEMORY.md | ask "Extract the key user preferences"

Generate JSON data:

bash

ask --json "List 5 sci-fi book titles and authors"

Complex Reasoning:

bash

ask -t "Solve this logic puzzle: Three gods A, B, and C are called..."