Coding & Development

Discover under-the-radar GitHub repositories using deep, multi-strategy topic graph searches. Use when the user asks to find hidden gems, underrated projects, missed opportunities, or non-obvious repos in a domain, technology, or topic. Sweeps topic clusters, star bands, recency windows, and low-star ranges, then ranks by gem signals (velocity, activity, README quality) rather than raw stars. Not for finding the most popular repo in a category, and not for code search.

Stops Claude Code from deleting your session history after 30 days by setting cleanupPeriodDays to 365 in ~/.claude/settings.json. Backs the file up first, keeps every other setting, and shows the result.

Picks one repeated job from section D of your Already Told You report and sets up its automation with you - a skill, a skill plus an app connection (MCP), or a scheduled run - always ending with one human approval step before anything is sent, published, paid or deleted.

>- Structural code search via ast-grep — use when code shape and element relationships matter, not just text. E.g., "find async functions without error handling", "refactor foo(a, b) to foo({ a, b })". Use Grep for simple name lookups.

Maintainer-only. Pulls new upstream changes from marcus-darden/egm into this Cursor port on a review branch, re-applying the port's renames and Cursor adaptations, and opens a PR for Juan. Run only when typed as /egm-sync-upstream.

装 skill 之前先查一句：这个值不值得装。它扫 skill 的目录，看声称的功能有没有对应的东西撑着——有能跑的代码，或者步骤清单写全的文档，算过；说要做某件事却一行实现都没有，标『别装』。也能盘点已装的 skill，指出哪些可以删。触发：值不值得装、这个skill能装吗、skill体检、装完有没有用、哪些skill可以删、skill是不是空的、skill review。

Turn a skill idea into a finished, tested, reviewed and released agent skill in one run - spec, tests first, an independent review against a release checklist, and a release that waits for the user's yes. Use for "skill factory", "/skill-factory", "build and ship a skill for X", "make this a skill and release it", or when loot hands over a drop whose form is a skill.

Builds or checks a simple skill with the user, for use in Claude, ChatGPT or Copilot.

根据业务需求生成或修订中文 MRD 需求内容，重点输出单界面的核心逻辑、明细列表和筛选项；适用于仲裁、理赔、工单、网点配置等业务系统需求。用户只要求讨论内容时不处理 Word 排版或界面图。

>- Checks that D source files and markdown documentation adhere to the project's 80-column line length rule. Run via rdmd.

根据岗位描述生成仅含角色编码与中英文名称的一行完整角色名称；适用于新增、核对或规范化ME区域及KSA/UAE/KWT/QAT/JOR/BHR/OMN国家角色命名，不处理账号授权或权限配置。

根据流程描述生成或核对OA流程编码与中英文完整名称，沿用ME-HR-001 Offer审批 | Employment Offer Approval格式；适用于流程命名，不处理角色命名、流程配置或审批发起。

Purpose: full scientometrics review of a manuscript **before submission to

Work with the user on their Margin drawing board (an iPad pen canvas). Use when the user mentions Margin, their board, drawing, sketch, page or whiteboard, says "check my board" or "look at my drawing", or asks you to answer, explain, plot or diagram something they drew.

Use when the user wants to reply to a code review remark on a GitHub pull request and resolve its thread. Triggers on "answer remark #N in the PR", "reply to the reviewer", or "resolve this review comment".

指导 AI 正确使用 PowPop MCP 连接器的 9 个工具：附近发现（数字人/用户/帖子）、话题搜索、数字人与帖子详情、景点商圈、活动与会员方案、身份检查。

Use when the user wants to see unresolved PR comments, review remarks, or outstanding feedback on a pull request.

Use when the user wants to respond to a code review remark by adding a clarifying comment to the source code — so the reviewer understands the reasoning and won't raise the same remark again. Triggers on "answer remark #N", "add comment to address reviewer concern", "document why we did X", or any request to resolve a review finding through in-code explanation rather than a code change.

Use Trellis channel for live multi-agent collaboration, spawned workers, cross-agent review, progress inspection, forum channels, and channel log debugging.

description: 当用户给出一个具体的 URL，要求从这个页面提取信息时使用。适用于单页内容提取、单页日期/数据抽取。不适用于用户只给出查询词、没有给出 URL 的情况——那种情况由 multi-source-extractor 处理。不适用于 GitHub 仓库解读——那种情况由 github-repo-explainer 处理。

Use this skill whenever the user asks to optimize code or make an application faster, more reliable under load, lower-latency, higher-throughput, or more resource-efficient. Trigger for profiling and performance audits; CPU or memory hot spots; slow endpoints; database queries, N+1 patterns, indexes, and round trips; or backend/API call latency, fan-out, and concurrency. Establish evidence, target the real bottleneck, and make the smallest maintainable change that improves the relevant outcome without weakening correctness, reliability, or YAGNI.

description:当用户给出一个查询词（而非具体 URL），要求从多个来源搜索并合并信息时使用。例如“帮我查一下某政策的最新消息”“找几个来源对比某个数据”。不适用于用户已经给出具体 URL 的情况——那种情况由 web-extractor 处理。

Extracts a GitHub URL (repo, issue, pull request, commit, file, release) into complete local artifacts AND explains in plain language what the project is trying to do, before presenting any details. Use when given a github.com link to "extract", "pull", "save", "read in full", or "explain", especially when the user says the result was unclear or when the URL carries a #number fragment.

Use when building C# applications with .NET 8+, ASP.NET Core APIs, or Blazor web apps. Builds REST APIs using minimal or controller-based routing, configures database access with Entity Framework Core, implements async patterns and cancellation, structures applications with CQRS via MediatR, and scaffolds Blazor components with state management. Invoke for C#, .NET, ASP.NET Core, Blazor, Entity Framework, EF Core, Minimal API, MAUI, SignalR.