Architecture Decision Records (ADRs)

← Home

All coding projects initialize the appropriate AI ops project

Introduction

To make AI-assisted work reliable and consistent across the team, every coding project needs the same baseline configuration: project instructions, code review skills, platform-specific rules, and shared prompts. We maintain this baseline in the ops project, claude-ops.

Standardizing on it means a dev moving between projects finds the same conventions, AI tooling behaves predictably, and improvements made in one place reach every site without per-project rework.

Decision

Every new coding project initializes the AI ops project that matches its stack:

Initialization uses the standard installation workflow so projects track the current version and pull updates centrally rather than copying configuration into each site. The ops project supplies CLAUDE.md instructions, code review skills, platform-specific rules, and the shared prompt and skills library.

Any deviation from the standard configuration is documented in the individual project.

Context

This is done for several reasons:

  1. A shared baseline makes AI-assisted development more reliable. Tools follow the same rules and review standards on every project.
  2. The self-updating mechanism lets us improve the ops project once and have every site benefit, instead of updating individual repos by hand.
  3. Devs moving between projects do not have to relearn conventions. Onboarding to a new project is faster.
  4. The ops project is where shared prompts, skills, and platform quirks live, so it is the natural home for organizational knowledge rather than per-project copies that drift apart.

Consequences

Exceptions

Additional Resources

None

← See more ADRs