Research & Data

Use when the user needs deterministic data transformation without an LLM — dedupe, money/date normalization, export diffing, secret scrubbing, provenance packets, or parsing Kindle/eBay/crypto exchange exports. Covers the punctual_alghoza Apify actor family.

Real-time transport firewall and boundary enforcer for Model Context Protocol (MCP) servers and Claude tool execution. Blocks Cloud Metadata SSRF (169.254.169.254), path traversal (../../), and destructive system commands.

Use when the user is new to GitHub or does not want to learn it and needs to publish files, create a repository, upload a folder, make a release with downloads, choose a licence, or remove private information from a public repo. Guides one click at a time and keeps private data out of public history.

ユーザーが調査実行の指示として「スキルを用いて」かつ「ローカル調査」または「ローカル検索」を明示的に記載した場合のみ起動する。この条件への単なる言及・引用、スキルの設計や改修に関する相談、通常の調査・検索・比較・要約・ファクト確認・質問への回答、類似表現では使用しない。起動時はローカルの skill-source ディレクトリとWeb検索の両方を使い、出典を明記した構造化レポートを1件生成する。

>- Activates the OWL (Optimized Workforce Learning) decoupled multi-agent pipeline: uses DeepInvestigator for root-cause analysis and edge-case attacks, and DeepCoder for verified code synthesis and iterative improvement.

Longitudinal statistical analysis pipeline. Use when the user has panel / repeated-measures data (one or more sessions per subject, with between-subject conditions) and wants to run a reproducible analysis: descriptive stats, a manipulation check, a linear mixed model, a lagged mediator, or a sensitivity analysis over random-effects / covariance structures. Triggers on any of the phrases "LMM", "mixed model", "repeated measures", "longitudinal analysis", "mediation over time", "lagged mediator", "sensitivity analysis of RE structure". When fired, load the YAML config that lists the enabled steps, run the pipeline, and report the per-step results plus the generated CSVs.

Cross-document conflict detector. Use when the user has two or more documents, datasets, or extracted fact tables and wants to find contradictions between them. Triggers on "detect conflicts", "find contradictions", "do these documents agree", "cross-check claims", "inconsistency check", or when comparing a claim across multiple sources. Runs three detection levels: entity (numeric / boolean mismatch on the same key), clause (categorical mismatch), and argument (free-text disagreement). Emits a machine-readable conflict report in JSON / CSV / Markdown.

Develop against ruoyi-vue-pro/Yudao projects, including backend modules, Vue3 admin UI, uni-app UI, SQL, permissions, and data-integration boundaries.

>- Provides foundational Firebase CLI setup, CLI installation, version checks (`firebase-tools@latest --version`), CLI login (including --no-localhost), project creation, project selection (`firebase use`), and app config file downloads (`google-services.json`, `GoogleService-Info.plist`). Use ONLY for CLI login, project creation/switching, or downloading app config files. Don't use for Firebase Hosting deploy, Firestore, Auth, App Hosting, Data Connect, Crashlytics, or Remote Config.

Research how far stocks or ETFs sit below their 52-week high, why they fell, and the bull and bear case, then publish a consistent HTML report and, if the user gives an email address, prepare a ready-to-send email of it. With no tickers named it uses a default watchlist (XLK, FTEC, QQQ, SMH, ARTY, S&P 500, Dow Jones). Use whenever the user lists tickers or asks for a market check, 52-week highs, drawdowns, "how much has X dropped", pullback analysis, or a drop/bull/bear report, even if they don't say "report".

Use Slashy MCP for connected Gmail or Outlook mail, Google or Outlook Calendar, contact research, meeting prep, reminders, and scheduled workflows. Covers connection setup, tool conventions, deep links, and attachments.

Use when code review feedback has arrived and needs a decision before any fix - a cr-request result, pull request comments, CI or static-analysis review output, or a reviewer's message - in any programming language. Verifies each finding against the current code, decides accept, reject, partial, defer, or escalate with evidence, fixes only what holds up, and reports per finding.

SQL-миграции PostgreSQL: парные up/down-файлы, современные типы и ограничения (identity, text с CHECK, timestamptz, jsonb), откат каждой операции, добавление значения в enum, индексы на живых таблицах через CREATE INDEX CONCURRENTLY без блокировки записи, что требовать от мигратора, разбор «индекс создан, а план — Seq Scan». PostgreSQL migrations: paired up/down files, modern types, reversible changes, non-blocking indexes, migrator requirements.

Как найти, что тормозит в PostgreSQL, и доказать, что правка помогла. Профиль по pg_stat_statements, сравнение вариантов по числу страниц (BUFFERS), а не по секундам, разбор плана от места, где оценка строк впервые разошлась с фактом. N+1 из Go-цикла сворачиваем в один запрос. Правим условия, FK, статистику и память, а не боремся с планировщиком. Postgres query tuning: profile first, compare by buffers, fix row estimates, remove N+1.

Import the user's ChatGPT or Claude data export into their AI file system, so it starts out knowing what they've already told AI. Use during setup, or when they say "import my ChatGPT history", "use my Claude export", or run /ai-file-system:import-history.

Runs a time-boxed research spike for one ticket of type spike (from the team's tracker such as Linear, or a local tasks file) and produces a decision record with evidence, a recommendation, and proposed backlog changes, without shipping code. Use only when a specific spike ticket (such as ENG-130 or T-03) is named, by the user or by the deliver-project orchestrator.

Comprehensive guide and development standards for maintaining, building, and extending the Campus Helpdesk React portal. Covers component architecture, responsive UI, issue validation, ticket lifecycle statuses, analytics, and verification workflows.

Orient in local codebases at the start of coding sessions and before refactors or complex code questions using scoped ast-grep outlines. Choose rg for exact text, ColGREP for behavioral discovery, and ast-grep for structural searches and transformations; not for web research.

>- Starts the post-commit reviewer for this project. One analysis, then a script-rendered Cursor canvas. Use when a coherent feature is done, or the user says review-change. Do not review in this chat. Do not launch this skill when applying or skipping canvas diffs.

把一篇学术论文变成一整套可汇报的中文优先材料——论文精读、技术与术语解析、Word 文档、汇报 PPT、引用脉络（前驱 + 后续）、可延伸的 idea。凡是用户提到「分析这篇论文」「论文汇报」「组会 pre」「文献汇报」「做成 PPT」「这篇论文讲了什么」「谁引用了这篇论文」「这篇论文还能怎么改进」，或者丢过来一个 PDF / arXiv 链接并要求讲解，都应当使用本 skill——即使用户没有明说「要一整套材料」。用户只说「帮我把这篇论文做成 PPT」时也要走本 skill 的完整流程：先精读、先核对，再动手做 PPT，避免整份 deck 建立在对论文的误读之上。PPT 采用白底、较大字号、图文分区和原生项目符号，版式主基准仍为 `references/group-meeting-sample.pptx`，每页含中南大学校徽。

Viết báo cáo theo một trong hai dạng: chuẩn quốc tế hoặc phổ biến tại Việt Nam. Tổng hợp nguồn có trích dẫn, giữ báo cáo chính gọn và tách dữ liệu hỗ trợ khi cần; xuất `.report.md` hoặc `.report.docx`.

Vẽ sơ đồ và biểu đồ đẹp cho báo cáo — flowchart/sequence/class/state/ER/gantt bằng Mermaid (kroki.io, PNG), biểu đồ cột/tròn/line bằng matplotlib (uv), output `<tên>.diagram.png` cạnh nguồn để nhúng vào `.report.docx`. Dùng khi báo cáo cần sơ đồ thay cho ASCII, cần biểu đồ dữ liệu, hoặc user yêu cầu "sơ đồ đẹp".

Trích xuất và chuyển đổi tài liệu PDF/DOCX/XLSX — lấy text layer (pdftotext), OCR ảnh scan (pdftoppm + tesseract), chuyển markdown (docling), XLSX → CSV (libreoffice headless), ghép/tách/đảo trang (qpdf/pdftk), kiểm tra output bằng pdfinfo + pdftotext. Dùng khi đưa tài liệu và muốn lấy text, OCR bản scan, chuyển định dạng, tách/ghép/trang, hoặc xác minh nội dung đã trích.

Reverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.