Agent 团队需要可重复的方法来衡量质量、回归和上下文变化。此技能指导评分标准设计、测试集、评分和监控。

当上下文窗口填满时，长时间 AI 会话会丢失关键决策、文件变更和后续步骤。此技能为 Claude、Codex 和 Claude Code 提供结构化压缩方法和评估探针。

构建认知代理的团队需要一种清晰的方法，将上下文映射为信念、欲望和意图。此技能提供本体模式、RDF 示例、SPARQL 检查以及集成指导，用于可追踪的 BDI 建模。

AI coding work can become overcomplicated, unfocused, or difficult to verify. This skill promotes explicit assumptions, minimal changes, and measurable success criteria.

Solana 团队会花时间连接 Anchor、测试和钱包前端。此技能会生成一个完整的入门项目，并提供一致的版本和设置步骤。

终端界面工作通常需要了解组件、布局、输入、动画和测试中与框架相关的 OpenTUI 细节。此技能为 Claude、Codex 和 Claude Code 提供结构化参考，用于构建和排查 OpenTUI 应用。

团队经常汇报活动进展，却没有把需要做出的决策讲清楚。Pitchcraft 可将原始项目事实转化为面向受众的简报，包含证据、风险和明确请求。

Manual web QA can be slow to repeat and difficult to observe across browser states. This skill helps you create, monitor, and manage Momentic Mo sessions with explicit test scope.

Browser tests can become fragile when steps, timing, and reusable flows are handled inconsistently. This skill guides structured Momentic authoring, validation, and troubleshooting.

Feature changes can drift from intended product behavior when implementation begins before end-to-end specifications. This skill defines focused Momentic tests first and validates durable checkpoints.

Test failures can hide their true cause behind later errors or retries. This skill investigates Momentic run evidence and produces a precise classification.

Mobile test authoring can fail when platform rules, device state, and Momentic formats are unclear. This skill guides focused creation, validation, and maintenance workflows.

Failing Momentic tests can hide the earliest broken state and waste repair time. This skill investigates evidence, classifies causes, and guides verified repairs.

Momentic explore runs fail when repository details, target URLs, authentication, and test paths are unclear. This skill produces a focused explore-prompt.md from verified repository facts.

ModelArts 团队需要针对 Notebook、训练作业、推理服务和云资源的安全、可重复协助。此技能会将请求路由到专门的 ModelArts 模块，并返回运维结果。

Choosing the correct mloda plugin pattern requires knowledge of its registry and extension points. This skill guides discovery, implementation, testing, packaging, and publication.

AI agents often generate fragile data-fetching and transformation code. This skill uses mloda feature requests, plugins, and lineage to build deterministic data workflows.

团队可能会通过移除来源署名、同意、来源记录或撤销机制来误用 Diasporic Intelligence。此技能提供清晰定义、署名措辞、边界测试和实施检查。

以图像为主的工作通常需要快速描述、OCR 和视觉审查。此技能会通过聚焦的提示词将图像路由到 MiniMax 视觉分析。

手动构建原生 PowerPoint 幻灯片可能既耗时又不一致。此技能为 Claude、Codex 和 Claude Code 提供精确的 PptxGenJS 规则，用于幻灯片布局、样式和 QA。