AgDR stands for Agent Decision Record. It is a specialized standard for documenting technical decisions made by AI coding agents, providing a structured way to track why specific implementation paths were chosen.

Where are the decision records stored?

The skill automatically creates and manages markdown files within a standardized directory structure at docs/agdr/ inside your project repository.

When should Claude use the agdr-decide skill?

It should be used for any non-trivial decision, such as selecting a new framework, choosing between design patterns, or resolving explicit 'X vs Y' comparisons before code implementation starts.

How does this help team collaboration?

It provides human developers with a clear 'paper trail' of the AI's reasoning, including the alternatives considered and the specific constraints that led to the final decision.

Can I skip the AgDR process for small changes?

Yes, the skill is designed to be skipped for trivial choices or when strictly following existing project conventions, focusing documentation efforts on high-impact decisions.

AgDR Decision Records

Name: AgDR Decision Records
Author: me2resh

byme2resh

•

Productivity & Workflow

Enforces a structured workflow for documenting technical decisions made by AI agents using the Agent Decision Record (AgDR) standard.

The AgDR Decision Records skill implements a rigorous governance framework for AI-assisted development by extending the traditional Architecture Decision Record (ADR) pattern. It forces a systematic approach to non-trivial technical choices—such as library selection, architectural patterns, and tool comparisons—ensuring that every significant decision is justified, documented, and stored within your repository. By requiring multi-option comparisons and 'Y-statement' justifications, it creates a transparent audit trail of the AI's logic, making project maintainability easier for human developers and future AI sessions alike.

Key Features

01Mandatory multi-option comparison tables to prevent biased decision-making

0225 GitHub stars

03Structured 'Y-statement' logic to capture context, goals, and tradeoffs

04Automated generation of zero-padded AgDR markdown files in docs/agdr/

05Automated sequencing and slug generation for consistent documentation

06Intelligent context gathering focused on performance, cost, and constraints

Use Cases

01Maintaining a historical record of technical debt and trade-offs accepted during development

02Documenting the rationale behind complex architectural or design pattern choices

03Evaluating and selecting between competing libraries or frameworks (e.g., 'React vs Vue')

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add me2resh/agent-decision-record agdr-decide

For use in Claude.ai and ChatGPT

Download Skill