Fabric: 200+ Reusable AI Prompts You Can Pipe From the Terminal
If you've ever found yourself rewriting the same AI prompt over and over—or trying to remember that one perfect phrasing for summarizing a meeting, extracting action items, or explaining a concept—you're not alone. Most of us end up with a messy notes file or a browser bookmark folder full of "good" prompts.
Fabric solves that in a surprisingly clean way. It's a terminal tool that gives you 200+ curated, reusable AI prompts you can pipe data into directly from your shell. No copy-pasting, no switching tabs, no "let me just tweak this one more time."
What It Does
Fabric is a command-line tool built around a simple idea: prompts are functions. You pipe text in, the prompt gets applied, and you get structured output back. The prompts cover everything from summarizing YouTube transcripts to generating git commit messages, extracting wisdom from articles, or even refactoring code.
Under the hood, it's a Go binary that wraps around an LLM API (like OpenAI). The prompts are stored as markdown files, so they're human-readable and easy to modify. You can also add your own prompts and share them with the community.
Why It's Cool
The real magic is the piping. Because Fabric works with standard Unix pipes, you can chain it with other tools effortlessly. For example:
youtube-transcript https://youtu.be/some-video | fabric --pattern summarize
Or grab a webpage, summarize it, then ask for action items:
curl https://example.com | fabric --pattern extract_wisdom
This isn't just about convenience. It's about making AI prompts feel like first-class citizens in your development workflow. You don't think "I need to open ChatGPT, paste this text, and manually select a prompt." You think "I need to pipe this through the 'create_summary' pattern."
Other cool features:
- Offline-first: The prompts are stored locally, so you can browse them without an API call.
- Pattern system: Each prompt is a "pattern" with a specific purpose. There are patterns for writing, coding, analysis, and more.
- Customizable: You can fork the repo, tweak prompts, and even submit your own back.
- No lock-in: It's just markdown files and API calls. You could swap out the LLM provider or even use it with a local model.
How to Try It
Getting started is straightforward. You'll need a Go compiler (or you can use the pre-built binaries) and an API key from OpenAI (or another supported provider).
Install via Go (if you have Go installed):
<