adr-skill

Create and maintain Architecture Decision Records with implementation plans and verification criteria.

27|16|Updated Apr 3, 2026
One-click install
npx skills add https://github.com/mandor1428/TEAM-LLM --skill adr-skill-mandor1428
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: adr-skill
Source: https://github.com/mandor1428/TEAM-LLM/tree/main/skills/adr-skill
Command: npx skills add https://github.com/mandor1428/TEAM-LLM --skill adr-skill-mandor1428

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

It helps teams capture architecture decisions in a way that AI coding agents can implement immediately, without needing follow-up questions or relying on tribal knowledge.

Core Features & Use Cases

  • Four-phase ADR workflow: scans the repo, captures intent via Socratic questioning, drafts the ADR, and reviews it against an agent-readiness checklist.
  • Self-contained, agent-first ADRs: enforces measurable constraints, explicit non-goals, and an implementation plan that names affected paths, dependencies, patterns to follow/avoid, configuration changes, and migration steps.
  • ADRs you can maintain over time: supports proposing, accepting/rejecting, deprecating, and superseding ADRs while linking decisions to the code they govern.

Quick Start

Ask a human to approve a proposed architectural choice, then run the skill workflow to scan existing ADRs, capture confirmed intent, draft the ADR with an implementation plan and verification checklist, and review it for agent readiness.

Frequently Asked Questions about adr-skill

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write architecture decision records for agentic coding workflows?

Architecture decision records for agentic coding require self-contained drafts with concrete implementation plans and agent-verifiable verification criteria, using templates and conventions to ensure AI agents can implement changes without follow-up questions.

What is the best way to maintain architecture decision records over time?

Maintaining architecture decision records involves supporting proposing, accepting, rejecting, deprecating, and superseding decisions while linking them to the code they govern, ensuring decisions remain traceable throughout the software lifecycle.

How do I create an implementation plan that AI coding agents can follow?

Creating an agent-ready implementation plan requires naming affected paths, dependencies, patterns to follow or avoid, configuration changes, and migration steps, ensuring the plan contains measurable constraints and explicit non-goals.

Can I use templates to enforce repo conventions for decision tracking?

Yes, applying provided templates and conventions ensures decision tracking aligns with repo standards, enforcing measurable constraints and verification checklists that maintain consistency across architecture decision records.

What should an agent-readiness review checklist include for ADRs?

An agent-readiness review checklist verifies that an ADR is self-contained, featuring explicit non-goals, an implementation plan naming affected paths and dependencies, and agent-verifiable verification criteria to prevent reliance on tribal knowledge.

Does this ADR workflow require scanning existing records before proposing changes?

Yes, the ADR workflow scans the repository for existing records before capturing confirmed intent, ensuring agents consult prior architecture decisions before proposing, drafting, or updating new changes.