What problem does it solve? Architecture weaknesses accumulate silently in a growing codebase, and ad-hoc refactoring lacks a disciplined loop: findings get lost, risky structural changes get applied without approval, and documentation drifts from the code. This Skill provides a bounded, user-invoked loop that analyzes the architecture, assesses weaknesses by tier, fixes the safe ones, and gates the risky ones on explicit user decisions. ## Core Features & Use Cases - Tiered weakness routing: Small weaknesses go to a domain implementer, substantial ones go through a designer -> plan gate -> user approval -> implementers -> verifier pipeline, and structural reworks are never auto-applied. - Fresh assessment every round: The architecture map and the ASSESSMENT.md findings are recomputed by separate capture skills each round, so the loop never edits or prunes the assessment itself. - Bounded and honest execution: A hard cap of 3 improve rounds, fresh-session resume blocks between rounds, green build/test baselines, and a structured final report with resolved, deferred, and proposed-decision fields. - Use Case: Point it at the Orders module of a .NET API; it maps the architecture, finds a repository leaking an EF type (small) and a grab-bag namespace (substantial), fixes both through the gated pipeline, flags a persistence-dependency inversion for your decision, and loops until satisfied. ## Quick Start Run /project-architecture-quality-loop on the Orders module to analyze its architecture and work through the fixable weaknesses.
