What problem does it solve? When turning an approved design document into code, AI agents tend to over-engineer, silently refactor unrelated code, or push through situations the design never covered. This Skill enforces a disciplined implementation workflow that keeps code changes aligned with the approved plan and makes "stop and ask" the default behavior. ## Core Features & Use Cases - Checklist-Driven Execution: Executes steps from a {slug}-checklist.yaml in strict order, marking each step done only after its exit signal is verified. - Scope Guardrails: Blocks out-of-plan changes, patch-branch hacks, and unauthorized new terminology, routing discoveries into follow-up issues instead. - Standardized Completion Report: Produces a fixed-format report listing changed files, functions, plan deviations, and acceptance-scenario evidence for user review. - Use Case: After a feature design doc is approved, tell the agent to start implementation; it validates the design and checklist, executes each step, and reports back for review before acceptance. ## Quick Start Tell the agent "方案确认了，开始实现" to have it validate the approved design and checklist, then implement the feature step by step with a final structured report.
