What problem does it solve? Engineering changes that span multiple files, add features, or migrate databases often lack a written record of intent, making it hard for reviewers and future maintainers to understand why and how work was decomposed. This Skill enforces creating a structured implementation plan in the issues/ directory before coding or opening a PR. ## Core Features & Use Cases - Plan necessity judgment: Distinguishes changes that need a plan (multi-file edits, new features, DB migrations) from trivial ones (typo fixes, config changes) so overhead is only added where it pays off. - Structured plan template: Provides a Markdown format covering acceptance criteria, alternatives considered, task decomposition by feature (not by layer), risk handling, rollback steps, and definition of done. - PR-time plan generation: When no plan exists at PR creation, it reconstructs one from conversation history, git log, and git diff. - Pre-refactoring decision support: A reference guide helps decide whether to restructure risky code (e.g., a 3700-line file touched by 9 of 10 tasks) in a separate PR before implementing. - Use Case: Before adding a filtering feature that touches models, services, and UI, generate issues/TASK-1234_add-filtering.md with acceptance criteria and per-feature tasks, then implement each task with failing tests first. ## Quick Start Ask the AI to create an implementation plan in the issues/ directory for the feature you are about to build before writing any code.
