Skip to content
Conduit

skill

oh-my-llm

Infuse your zsh with AI

yvbbrjdr/oh-my-llmcommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

The download comes from github.com/yvbbrjdr/oh-my-llm itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 22, 2026

oh-my-llm

Infuse your zsh with AI

oh-my-llm is a zsh plugin that brings a large language model directly into your shell. Type natural language instead of commands and run web searches and fetches, all without leaving your terminal.

Features

  • Natural language commands: type a question or instruction at the prompt and the AI responds or runs the right command for you
  • Tool use: the AI can execute shell commands (with your approval), search the web, and fetch web pages
  • Skills: extend the AI with reusable, project- or machine-scoped capabilities loaded from local SKILL.md files
  • Configurable backend: works with OpenAI or any OpenAI-compatible API endpoint

Requirements

  • zsh
  • Python 3.8+ with venv and pip
  • An OpenAI API key (or a compatible provider)
  • (Optional) An Exa API key for web search and page fetching

Installation

With oh-my-zsh

1. Clone the repository into your oh-my-zsh custom plugins directory:

git clone https://github.com/yvbbrjdr/oh-my-llm ${ZSH_CUSTOM:-~/.oh-my-zsh/custom}/plugins/oh-my-llm

2. Add oh-my-llm to the plugins list in your ~/.zshrc:

plugins=(... oh-my-llm)

3. Reload your shell:

omz reload

Without oh-my-zsh

1. Clone the repository anywhere you like:

git clone https://github.com/yvbbrjdr/oh-my-llm ~/.oh-my-llm

2. Source the plugin from your ~/.zshrc:

source ~/.oh-my-llm/oh-my-llm.plugin.zsh

3. Reload your shell:

source ~/.zshrc

The plugin will automatically create a virtual environment and install its dependencies on first load.

4. Configure your API keys:

oml config

This opens a JSON config file in your $EDITOR. Fill in your credentials.

Usage

Natural language at the prompt

Just type what you want. If it isn't a valid shell command, oh-my-llm handles it:

$ what is my public IP address?
$ find all PDFs modified in the last week
$ summarize the contents of README.md
$ convert photo.heic to JPEG
$ convert all PNGs in this folder to WebP
$ compress video.mov to an MP4 under 50MB
$ extract the audio from recording.mp4 as an MP3
$ what does the -r flag do in rsync?
$ who owns the process on port 8080?

Session management

Conversation context persists for your entire shell session.

To clear the history and start fresh:

oml clear

To upgrade oh-my-llm to the latest version:

oml upgrade

This pulls the latest changes from git and reloads your shell.

Skills

Skills extend oh-my-llm with reusable, task-specific instructions the AI reads before tackling certain problems. A skill is a SKILL.md file with a YAML frontmatter header followed by freeform guidance. See agentskills.io for details.

Skill discovery

oh-my-llm automatically discovers skills from two locations:

  • Project skills: any skills/*/SKILL.md path found inside a hidden directory (e.g. .oml/skills/) under your current working directory, useful for per-project capabilities committed alongside your code.
  • Global skills: skills/*/SKILL.md files inside the oh-my-llm plugin directory itself, available in every session regardless of where you are.

Security

Shell commands proposed by the AI require your explicit approval before execution. When skip_readonly_command_verification is enabled, read-only commands (e.g. ls, cat, grep) are confirmed automatically and run without a prompt; commands that modify the system always require approval.

License

BSD 3-Clause. See LICENSE.

↑↓ navigate openesc close