| name | architecture-deepening-review |
| description | Use when reviewing a codebase for architectural friction, shallow modules, poor test seams, scattered behavior, duplicated orchestration, or opportunities to deepen modules before implementing a ticket or workflow. |
Architecture Deepening Review
Use this to find high-leverage refactors before a Decisions ticket or workflow implementation.
Workflow
- Read project domain language first. Use
domain-modeling if terms are unresolved.
- Use
codebase-design vocabulary: module, interface, seam, adapter, depth, leverage, locality.
- Explore where understanding one behavior requires bouncing across many files or where tests cannot reach real behavior through a public interface.
- Produce a short report with candidates, files involved, current friction, proposed deeper module, test improvement, and recommendation strength.
- Ask the user or workflow owner to choose a candidate before designing the new interface.
Report Rules
- Do not propose implementation during the first review.
- Do not fight existing ADRs unless the friction is concrete.
- Keep AgentManager patterns to assessment notes unless they preserve DecisionsAI one-agent-per-board orchestration.
- Prefer a local HTML or Markdown artifact outside runtime-critical paths.