Scan, Fix & Verify Code Quality Across All Dimensions

This skill runs a unified full-sweep analysis across all quality dimensions - code decay, architecture, tech debt, and test quality - then applies fixes directly to the codebase: safe changes are auto-applied, risky changes require confirmation before execution. It draws on the same shared framework as sibling brooks-lint skills - the Iron Law and Health Score rules, book-level source coverage, and separate symptom definitions for production risks and test risks - plus its own sweep guide for the unified scan-and-fix process. If no project or directory is specified, it applies automatic scope detection to determine the review scope first.

The process starts by showing a pre-flight consent notice and waiting for one-time user approval, then enumerates scope and initializes tracking state for unresolvable items, non-critical-round counts, and a fix log. It runs four dimensions in sequence - review, test, debt, and audit - each scanning, classifying, applying safe and extended-safe fixes, and verifying via the project's own test command. It then iterates by re-scanning modified files, the same module, and static consumers, converging on a clean round, retiring items that fail three retries to an unresolvable set, and capping non-critical rounds at three. Finally it aggregates residual and unresolvable items into a Full Sweep Report.

Use it for a comprehensive, auto-fixing pass across code decay, architecture, tech debt, and test quality in one run - not for a single-dimension check, where the dedicated audit, review, or test skills are the better fit.

Given a project or an auto-detected scope, and only after explicit user consent, it produces directly applied safe fixes, confirmed risky fixes, a fix log, and a Full Sweep Report listing residual and unresolvable items.

Depends on shared framework files defining the Iron Law, source coverage, and separate production and test risk symptom taxonomies, plus the project's own test command for fix verification.

Teams wanting a single, consent-gated pass that both diagnoses and fixes maintainability, architecture, tech-debt, and test-quality issues across a codebase, rather than running separate audit, review, and test skills one at a time.