skills/vibe-to-arch/SKILL.md
A full project lifecycle methodology covering two phases: fast exploration (vibe coding) followed by systematic rebuild (clean architecture). Use this skill whenever the user mentions starting a new project, wants to explore an idea without committing to architecture, says "MVP is done" or "validation complete" and is thinking about a refactor or rebuild, asks how to approach a large-scale refactor, wants to transition from messy prototype to clean system, mentions "vibe coding" or exploratory development, or asks about system design after building a working prototype. This skill applies even when the user doesn't explicitly ask for it — if they're showing signs of being between exploration and architecture, step in.
npx skillsauth add jasong98/code-skills vibe-to-archInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Architecture-first development is a bet on knowing what you're building before you've built it. That bet rarely pays off. The better model:
Phase 1 — Explore: Vibe code your way to understanding. Ship fast, break things, learn faster. The code is a side effect; the real output is knowledge.
Phase 2 — Graduate: When exploration converges, stop. Distill what you learned. Don't carry vibe-code assumptions into the new system.
Phase 3 — Rebuild: Now that you know what you're building, build it right. First principles → clean architecture → systematic implementation.
The key insight: vibe code is the best requirements document ever written. It's concrete, executable proof of what the system actually needs to do. The rebuild isn't a failure — it's the payoff for good exploration.
Before doing anything, identify which phase the user is in.
Ask (or infer from context) two questions:
| Situation | Go to |
|-----------|-------|
| Starting fresh / exploring an idea | phase1-explore.md |
| Prototype works, considering cleanup | phase2-graduate.md (don't skip this) |
| Graduation confirmed, ready to rebuild | phase3-rebuild.md |
| Both questions answered "yes" | phase2-graduate.md → phase3-rebuild.md |
| Either question answered "no" | phase1-explore.md |
Never jump from Phase 1 directly to Phase 3. Phase 2 exists specifically to prevent carrying vibe-code assumptions into the new system.
Load only the file for the current phase — not all three upfront:
phase1-explore.md — Strategic exploration: how to vibe code with intentphase2-graduate.md — Graduation: assessing readiness + spec extractionphase3-rebuild.md — Rebuild: first principles → architecture → implementation1. Wrong fast beats right slow in Phase 1. Premature correctness is just procrastination with good posture.
2. Knowledge has an expiration date. The gap between Phase 1 and Phase 3 should be days, not months. Context decays. LEARNINGS.md exists to fight this.
3. Spec extraction before architecture. Phase 3 starts by distilling what Phase 1 proved — not by designing from scratch. The vibe code is a research artifact, not a scaffold to build on.
4. Stop points are mandatory in Phase 3. This is a collaborative process. Claude surfaces decisions; humans confirm them. No unilateral architecture choices, no silent workarounds.
5. Strategic messiness. Not all Phase 1 code deserves equal care. Core logic and data models: leave traces. Scripts, UI, experiments: throw away freely.
6. Rebuilds can be partial. Not every refactor needs to be a full teardown. A single module can be rebuilt while others remain. This skill applies at any granularity.
development
使用 uv 工具管理 Python 项目的依赖、环境和工作流。适用于项目初始化、依赖管理、环境配置和 CI/CD 设置场景。
development
Python 代码编写规范和最佳实践。适用于代码生成、代码审查、代码重构场景。
development
根据今日 Claude Code 历史对话生成日报。输出两个版本:详细版(供个人阅读复盘)和简洁版(可直接提交)。日报追加写入桌面"日报"文件夹,每周一个 md 文件。当用户说"生成日报"、"写日报"、"今日总结"、"daily report"、"做日报"时触发此技能。
development
Maintainer-only workflow for handling GitHub Secret Scanning alerts on OpenClaw. Use when Codex needs to triage, redact, clean up, and resolve secret leakage found in issue comments, issue bodies, PR comments, or other GitHub content.