udecode/dotai

What it solves

It provides a centralized, reusable library of "skills" (workflows) for coding agents, preventing the need to copy long configuration files across multiple repositories.

How it works

The project acts as a canonical source for agent workflows. These skills serve as a routing layer that allows agents to perform specific tasks. Examples include autogoal for goal lifecycle management, tdd for test-first development, and resolve-pr-feedback for handling GitHub PR reviews. Users can install these skills using a dedicated CLI tool (npx skills add udecode/dotai).

Who it’s for

Developers building or using AI coding agents that require standardized, reusable operational workflows for software development tasks.

Highlights

• Goal Management: Includes autogoal for durable goal lifecycles and plan templates.
• Agent-Native Review: A specialized reviewer to ensure agents can perform the same actions as human maintainers.
• Development Workflows: Built-in support for TDD (Test-Driven Development) and PR feedback resolution.
• Infrastructure Integration: Integrates with tools like GitHub, Linear, Git worktrees, and Gemini API for video transcripts.