Manual contract emission interrupts development and can leave production artifacts outdated. This skill configures supported Vite automation and explicit build steps for other tools.

Vague Prisma Next questions can lead to the wrong workflow. This skill identifies the user goal and routes it to the appropriate specialized skill.

Prisma MongoDB 项目无法升级到 Prisma ORM v7，因为该版本不包含 MongoDB 连接器。此技能可帮助团队评估阻碍因素、安全地继续使用 v6，或规划分阶段的 Prisma Next 评估。

Prisma v7 驱动适配器对事务、值和错误有严格的契约要求。本指南为 SQL 适配器开发提供实现模式和验证清单。

为新数据库设置 Prisma 时，需要处理因提供商而异的数据源配置、驱动适配器和连接字符串。此技能为 Claude 提供针对特定提供商的分步指南，使其能够一次性正确配置 PostgreSQL、MySQL、SQLite、MongoDB、SQL Server 或 CockroachDB。

Prisma Compute 部署需要正确的配置、身份验证、环境设置和运行时端口。此技能针对受支持的 TypeScript 框架提供设置、部署、验证和故障排除指导。

Prisma Client 任务可能需要许多查询选项、关系模式和事务选择。本参考可帮助你选择已文档化的 API，并更安全地使用原始 SQL。

Prisma CLI 工作流可能会影响 schema、生成的客户端和数据库数据。此技能为每项 Prisma CLI 任务提供命令指导、安全边界和参考资料。

Hotel rates and cancellation terms vary across booking providers. This skill compares live results from supported sources and identifies the lowest available rate.

Comparing hotel prices across several booking sites is slow and inconsistent. This skill searches live sources and presents the cheapest dated options in one view.

Checking several hotel sites takes time and produces inconsistent currencies. This skill searches four sources and presents comparable nightly rates with booking links.

Flight prices vary across travel sites and can be difficult to compare consistently. This skill searches PriceWin data and presents ranked fares with booking links.

Comparing rooms and completing a booking across providers requires several careful steps. This skill guides discovery, confirmation, payment links, status checks, and cancellation.

AI 媒体工作通常需要分别使用生成、编辑和字幕工具。PonyFlash 将 SDK 指南与 FFmpeg 工作流结合起来，支持图像、视频、语音、音乐和导出。

插件作者在添加 commands、agents、skills、hooks 或 MCP servers 之前，需要一个可靠的结构。此技能说明预期布局、manifest 字段、命名规则和可移植路径模式。

插件作者需要一种安全方式来存储按项目的设置，同时避免提交用户特定状态。此技能说明了适用于 Claude、Codex 和 Claude Code 插件的本地 Markdown 与 YAML frontmatter 模式。

Manual plugin setup can produce invalid manifests and inconsistent marketplace entries. This skill scaffolds validated local plugins and manages personal marketplace metadata.

模型团队需要可重复的证据来衡量 agent 性能。PinchBench 运行结构化的 OpenClaw 任务，并生成可比较的分数，用于分析和排行榜提交。

Codex guidance becomes stale, duplicated, and difficult to trace. This skill maintains a review-first wiki with source provenance, explicit knowledge states, safe paths, and deterministic linting.

Repository changes can fail when implementation, testing, and review are disconnected. This skill manages a complete local engineering loop with explicit evidence and stop conditions.