WORK STORY / 01
Turn engineering standards from shared understanding into executable gates
Context
As the project continued to develop and be refactored, inconsistent module boundaries, scattered styles, static text, oversized components, and differing development approaches began to coexist. Once AI accelerated implementation, these inconsistencies could also be copied into new code more quickly.
My Decision & Action
I gradually moved objectively enforceable engineering rules into the toolchain: ESLint, Stylelint, pre-commit checks, and CI gates constrained code structure; i18n, Design Token, and Module Layering standards were established; and large components and testing responsibilities were separated. Shared instructions, Skills, and automated checks were also added to the Agent workflow so that people and AI followed the same engineering boundaries.
Outcome
Architecture and quality requirements no longer existed only in documentation or code review; they could be checked continuously during development, commit, and CI. Whether code was produced by an engineer or with AI assistance, it had to pass the same project rules.
Reflection
When AI can rapidly generate large amounts of code, what truly needs to scale is not review headcount, but the conversion of assessable engineering principles into machine-executable rules that stop errors before they enter the codebase.