Software Development Discipline and Code Quality Guard

Problem

AI-assisted coding risks are often not syntax errors, but misunderstood intent, out-of-scope edits, and surface patches. A bug fix may change unrelated functions, a feature request may skip confirmation before touching core logic, and the result is hard to review or roll back. This skill turns code changes into a confirmable, verifiable, and traceable engineering workflow.

How It Works and Where It Applies

It works around collaboration rules and code quality:
- Confirm before executing: restate the request, list affected files, break the work into tasks, and flag risks; no code is written before user approval.
- Control edit boundaries: change only in-task code and required follow-on fixes; record extra issues instead of refactoring casually.
- Back up critical changes: payments, orders, auth, schema/data changes, or multi-file edits should be backed up under _backup/date-task/.
- Prefer root causes: trace the call chain, avoid masking abstraction errors with patches, and extract shared logic when duplication appears.
- Stay consistent: follow existing naming, error handling, and security patterns, and avoid adding dependencies without approval.

Use it for coding, bug fixes, feature work, refactoring, and source file changes. Single-line fixes, pure config values, urgent fix-first incidents, or explicit "just change it" requests may relax confirmation, but minimal edits and a REVIEW self-check are still expected.