Skip to content
← Tools / Skills
ToolDaniel Miessler/10 trustVerified
Fabric logo

Fabric

Modular AI workflows from the command line — hundreds of crowdsourced prompt patterns for real tasks.

Visit official site →

About

Fabric is Daniel Miessler's open-source framework for augmenting humans with AI, and it solves the integration problem rather than the capability problem. The core unit is the Pattern: a versioned, Markdown-based prompt template for a specific real-world task — summarize an article, extract wisdom from a YouTube video, analyze claims, threat-model an API, rate content. Hundreds of them are crowdsourced in the repo, synced via git, and they run through one Go CLI against any supported model provider.

The workflow is aggressively Unix: pbpaste | fabric --pattern summarize, fabric -y <youtube-url> --pattern extract_wisdom, chain patterns with pipes. Patterns appear as models (summarize:latest), there is a REST API and web UI alongside the CLI, and the whole thing is MIT-licensed with 44,000 stars and daily commits. It is not an agent framework — there is no tool loop or memory here — but as a prompt-as-package-manager for repeatable knowledge work, nothing else has the pattern library.

What it can do

  • Hundreds of crowdsourced, versioned prompt Patterns for real tasks
  • One CLI against any model provider (plus REST API and web UI)
  • Unix-native: pipes, stdin, YouTube/URL ingestion built in
  • Patterns as Markdown — readable, editable, git-synced
  • Custom patterns: drop a .md file in `~/.config/fabric/patterns/`
  • Streaming output and pattern chaining for multi-step workflows

Setup

  1. Install the Go CLI (one-line installer script, Nix, or go install) — see the repo README.
  2. Configure your model provider API key (fabric --setup walks you through it).
  3. Sync the pattern library: patterns live in ~/.config/fabric/patterns/.
  4. Run one: pbpaste | fabric --pattern summarize or fabric -y <youtube-url> --pattern extract_wisdom.
  5. Write your own patterns as Markdown files in the patterns directory; share via git.

Prerequisites

A model provider API key (OpenAI, Anthropic, Ollama, and others supported); Go toolchain only if building from source.

Use cases

productivitydeveloper-toolsterminalopen-sourcewritingresearch

Works with

Trust score — 9/10

/10 trustVerified

Scored 9/10 (verified tier, metrics-v1). Evidence: 44,278 GitHub stars (GitHub API), MIT license, pushed Oct 15 2026; an independent distilled-knowledge review corroborates the pattern model, the vendor abstraction, and the CLI/API/UI distribution. Strengths: unmatched crowdsourced pattern library, genuinely open, daily maintenance, vendor-agnostic. Docked for: (1) it is prompt patterns, not agents — no tool loop or memory, so scope is bounded; (2) pattern quality varies since they are crowdsourced. License: MIT.

★ 44,278 GitHub stars

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
9.0/10
Maintenance · recency, releases, issue load
10.0/10
Openness · license + docs
10.0/10
Manual review · sanity read + red-flag check
8.0/10
  • · Adoption 9: starScore from 44,278 stars on danielmiessler/fabric
  • · Maintenance 10: pushed 2026-10-15 (recency 10); daily commits
  • · Openness 10: MIT + README verified
  • · Manual 8: independent distilled-knowledge review corroborates the pattern model, vendor abstraction, and CLI/API/UI distribution; docked for bounded scope (prompt patterns, not agents) and variable crowdsourced pattern quality

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.