quality-playbook

Quality Playbook runs a complete quality-engineering audit on any codebase in any language. It derives behavioral requirements directly from the code, generates spec-traced functional tests, runs a three-pass code review with regression tests for every confirmed bug, executes a multi-model spec audit (the "Council of Three" independent AI auditors), and produces a consolidated bug report with TDD-verified patches. Its stated aim is to catch the roughly 35% of real defects that structural code review alone misses, by tying every bug back to a requirement and every requirement back to an exploration finding.

The workflow is organized into phases with strict dependencies: prior-run analysis, exploration (written to EXPLORATION.md), artifact generation (requirements, constitution, functional tests, review and audit protocols), three-pass code review, spec audit with semantic citation checks, reconciliation with a TDD red-green cycle, self-check verification, and an interactive present/improve phase. An orchestrator agent spawns clean sub-agent contexts per phase, verifies output files exist on disk before proceeding, and reports progress; it explicitly authorizes writing .md files, patches, and JSON to a quality/ directory and the project root. The skill supports multiple hosts (Claude Code, Copilot, Cursor, Continue) and can ingest reference documentation to raise finding confidence.

Target users are quality and platform engineers, reviewers, and teams who want a rigorous, spec-traced audit rather than a surface code scan. It works best with supporting documentation available and is suited to auditing modules or whole codebases with drill-down and iteration options.