What problem does it solve? AI agents produce code faster than anyone can review it, and quality standards that live only in conversation get quietly lowered — tests skipped, checks suppressed, thresholds edited down. This Skill turns a project's quality bar into a written, versioned contract with real numbers and tooling, then watches diffs for attempts to weaken it. ## Core Features & Use Cases - Constraint Interview & Defaults: Runs a four-question intake with sane defaults (coverage, security scanning, performance budgets, accessibility) so even "I don't know" answers produce a working configuration. - CONSTRAINTS.md Generation: Writes a single repo-root contract with an always-enforced floor (no new suppressions, stubs, or skipped tests), numbered enforced dimensions with named tools and commands, measured-only ratchets, and expiring exceptions. - Floor Guard Reference: Ships a diff-scoped reference implementation that detects weakened thresholds, removed assertions, new suppression comments, and unimplemented stubs, exiting 0/1/2 for CI integration. - Use Case: Before running an autonomous build loop on a TypeScript project, use this Skill to record an 80% changed-line coverage gate, gitleaks and Semgrep security checks, and a Lighthouse LCP budget, then block any diff that lowers those bars. ## Quick Start Ask the agent to set up quality constraints for this project and write them into a CONSTRAINTS.md file with enforced thresholds.
