菜单
Use when planning or performing Lux backups, restore drills, persistence or TTL configuration, secure deployment, storage-mode decisions, or upgrades.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/lux-db/lux --skill recovery-testing命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Use when implementing or debugging Lux signup, login, sessions, OAuth, SSR cookies, project-key boundaries, providers, or row-level grants.
Use when setting up a Lux app, designing tables, writing CRUD or query code, choosing SDK versus HTTP or RESP access, generating types, or using local development.
Use when authoring, planning, applying, pulling, rolling out, or repairing Lux application migrations.
基于 SOC 职业分类
正在显示 SKILL.md
| name | recovery-testing |
| description | Use when planning or performing Lux backups, restore drills, persistence or TTL configuration, secure deployment, storage-mode decisions, or upgrades. |
Start with the requested operation's minimum safe runbook. Use only public README.md, DURABILITY.md, SECURITY.md, cli/README.md, and installed public CLI help. Do not read, cite, or derive consumer guidance from source, tests, engine internals, or undocumented settings unless the user explicitly asks to contribute to Lux. If public docs conflict, name it and prefer DURABILITY.md for durability decisions.
Record the Lux version, storage mode, configuration, and backup time. Create an authenticated snapshot or run SAVE, then protect the produced snapshot and metadata with the same access controls as application data.
Practice recovery in a fresh instance: stop writes, restore the snapshot, restart Lux, and verify INFO plus application-level invariants. Treat tiered-mode WAL recovery as bounded by the documented fsync window; memory mode is snapshot-only. Do not assume a multi-write Lua script is crash-atomic.
Before an upgrade, create and verify a backup, read relevant release notes, and use explicit lux update commands. Do not downgrade across a storage-format change without documented support. Keep RESP/HTTP and snapshot/restore endpoints private or authenticated; use --bind only for a deliberate trusted-network deployment.
