Set Clear Quality Guardrails for AI-Assisted Coding

AI coding tools can produce substantial work quickly, but they can also create oversized controllers, repeated interface code, or unsuitable queries when they lack project context. Output quality depends heavily on the information and boundaries they receive.

Describe the repository structure, component patterns, localization rules, layer boundaries, and performance and security requirements. Give the tool examples from the existing code and define acceptance criteria that can be verified. Prefer small, reviewable changes over assigning a broad surface in one pass.

Treat the output as a proposal under full human responsibility. Review the diff, run analysis and tests, inspect failure cases, security impact, and performance, and do not expose secrets or sensitive data without a clear policy. The tool multiplies team capability inside a sound engineering process; it can multiply disorder when that process is absent.