I made a simple harness for AI-assisted coding (github.com)

A new open-source tool, Quality Workflow Meta, addresses a critical issue in AI-assisted coding: the tendency for AI-generated code to become complex, tightly coupled, and difficult to maintain. By automating complexity limits, linting, testing, and continuous integration (CI) gates via Git hooks, this harness forces AI agents to meet stringent quality standards before commits or pushes are allowed. This prevents technical debt accumulation by ensuring code remains modular, well-tested, and compliant with style and complexity metrics throughout development. Technically, the tool supports JavaScript/TypeScript (with Vite/Vitest) and Python stacks, integrating linters like ESLint and complexity checkers such as FTA, Xenon, and Radon. It sets up local pre-commit hooks (via Husky or pre-commit) and CI workflows that block commits failing tests or exceeding complexity thresholds. The installer is ephemeral, bootstrapping hooks, configs, and docs while self-destructing to keep repos clean. Developers follow a strict "Agent Working Agreement" to prevent bypassing checks and enforce incremental, transparent code changes. The solution offers a practical baseline for teams relying on AI coding tools, improving reliability and maintainability of AI-assisted software projects by baking quality assurances directly into the workflow.