📄 中文摘要
在为 G-Kentei 考试学习的过程中，开发了一款基于间隔重复的测验应用程序，分别使用 Python/Streamlit 和 Swift/SwiftUI 创建了 Web 版和 iOS 版。通过测试驱动开发（TDD）方法，在 Everything Claude Code (ECC) 环境中实施，记录设计决策并形成了有效的软件。此工具名为 pdf2anki，旨在解决内容而非应用程序本身的问题，帮助用户更有效地进行学习和记忆。
永远不要信任大型语言模型的输出 — 从构建 PDF 到 Anki 的 CLI 的六种防御措施
出处: Never Trust LLM Output — 6 Defenses from Building a PDF-to-Anki CLI
发布: 2026年2月20日
📄 English Summary
Never Trust LLM Output — 6 Defenses from Building a PDF-to-Anki CLI
A quiz app based on spaced repetition was developed to study for the G-Kentei exam, featuring both a web version in Python/Streamlit and an iOS version in Swift/SwiftUI. Implemented using Test-Driven Development (TDD) in an Everything Claude Code (ECC) environment, design decisions were recorded in Architecture Decision Records (ADRs), resulting in functional software. The tool, named pdf2anki, addresses the issue of content rather than the application itself, aiding users in more effective learning and memorization.