What problem does it solve? AI coding agents produce more code than anyone reviews, and without a written quality bar they quietly lower standards — adding suppressions, skipping tests, or weakening thresholds to get checks passing. This skill turns your project's quality bar into an explicit, versioned contract that outlives any single session. ## Core Features & Use Cases - Constraint Interview & Defaults: Runs a four-question intake with sane defaults for coverage, security, performance, and accessibility dimensions, then records decisions in a CONSTRAINTS.md file at the repo root. - Tool-Backed Enforcement: Maps each quality dimension to a de facto tool (tsc, Semgrep, gitleaks, osv-scanner, Lighthouse, axe-core, Stryker) with lifecycle placement so fast checks run per-edit and slow ones run in CI. - Floor Guard Reference: Ships a diff-scoped reference implementation that detects the five bar-lowering moves — weakened thresholds, skipped tests, new suppressions, unimplemented stubs, and untracked exceptions. - Use Case: Before running an autonomous build loop on a TypeScript project, use this skill to establish an 80% changed-line coverage gate, secret scanning, and a ratchet that holds today's project coverage, so the agent cannot ship code that silently degrades quality. ## Quick Start Ask the agent to set up quality constraints for this project and write them into a CONSTRAINTS.md file with enforced thresholds.
