Fabriq

Key Concepts

Understand the ideas behind Fabriq's design — the mental model you need before using the product.

Before you dive into the step-by-step guides, it helps to understand a few foundational ideas. Fabriq is built on a simple principle: AI agents should be powerful but never invisible. Every design decision flows from that.

These concepts appear throughout the docs. Skim them now, then come back when you encounter a term you don't recognize.

The Two Surfaces

Fabriq separates your work into two distinct surfaces:

  • Workspace assets — long-lived, shared product knowledge (docs, specs, ADRs, design tokens)
  • Feature worktrees — short-lived, isolated branches where the agent does its work

The shell enforces this boundary. The agent can never accidentally write across it.

Read more: Workspace vs Feature: Two Surfaces

Confirmation Gates

Every state-changing action the agent takes surfaces as a confirmation prompt — you see the target file, the dimension, the payload, and the expected effect. Nothing hits disk until you approve.

Read more: Confirmation Gates

Stages & Gates

A feature progresses through stages — from analysis through design, implementation, validation, review, and close. Between each stage is a gate: a deterministic check that ensures prerequisites are met before you can advance.

Read more: Feature Lifecycle · Stage Gates

The Audit Trail

Every capability call, every confirmation, every stage advancement is logged in an append-only JSONL audit trail. Every agent decision leaves sourced, attributable evidence.

Read more: Audit Trail

CDR (Cognitive Derivation & Reverse Engineering)

Fabriq doesn't just manage features — it learns from your codebase. CDR scans your source code and extracts structured knowledge: behaviors, business rules, state machines, and evidence that both humans and agents can navigate.

Read more: CDR Overview

Next steps

Edit on GitHub

Last updated on

On this page