AI & ML

leftzzzz Bundle

Audit, reorganize, split, migrate, and maintain AGENTS.md, optional platform instruction files, and current-only engineering Specs with cross-agent routing and SDD-aligned docs/specs, docs/requirements, and docs/technical directories. Use when cleaning up agent instructions, organizing engineering standards, contracts, policies, or invariants, removing history or summaries from Specs, separating business requirements and technical decisions from Specs, migrating rules out of .agents, checking multi-agent compatibility, or installing deterministic structure guardrails.

quantumedu Bundle

Trigger: /q-skillvault, /skillvault, skillvault, q-skillvault, vault de skills, knowledge vault, base de conocimiento, line factory, mcp skillvault, context bundle. Knowledge Operating System local-first para desarrolladores y agentes IA: almacenamiento híbrido (SQLite+FTS5), compilación contextual de 7 modos, MCP stdio server, router semántico y fábrica determinista issue-to-PR (line).

sepinetam Bundle

使用 cnki-mcp 检索知网文献、按篇名和作者等字段组合查询、获取摘要及题录元数据、列出期刊某一期目录。适用于知网文献检索与元数据整理；Chrome 或 computer use 网页操作使用 explore-cnki-view。

sepinetam Bundle

使用 Chrome 或 computer use 操作知网页面，完成网页检索、文献详情与期刊目录浏览、题录摘要和引用信息提取，以及经用户确认的页面下载。用户明确要求浏览器操作或需核对真实网页时使用；cnki-mcp 文献检索调用由 cnki-search skill 负责。

adibaarooj Bundle

Act as a stock market trading agent that fetches live market data, technical indicators, and TradingView technical-analysis signals for stocks, ETFs, forex, crypto, indices, and commodities, then produces a structured trading analysis with a directional recommendation, entry/exit levels, stop-loss, risk assessment, and position-size guidance. Use this skill whenever the user asks for a market/trading analysis, trade signals, buy/sell recommendations, technical analysis, price targets, symbol screening, or anything that mentions TradingView, tickers, candles, OHLCV, RSI/MACD/EMA indicators, or "analyze this stock/coin/forex pair". Also triggered by "stock market trading agent", "trading bot analysis", or "signal generator". The agent is analysis-and-signal only — it does not execute trades; it still returns honest, data-backed analysis when live TradingView data cannot be fetched, clearly flagging any estimates.

adibaarooj Bundle

Execute and investigate banking test cases, then produce a standardized, evidence-backed Observation Package for a downstream RCA system. Use this skill when the user provides a banking test case (ID, steps, expected result, environment, etc.) and needs investigation results — test execution, evidence collection, expected-vs-actual comparison, and a machine-readable observation JSON. Covers UI and API banking tests across accounts, transactions, payments, compliance, cards, lending, and core banking. This skill is for upstream investigation only — it never makes RCA classifications or defect decisions. Trigger on: banking test, test observation, test investigation, test evidence package, RCA input, test discrepancy, banking QA, payment test, account test, card test, compliance test, loan test, transaction test.

m1nga Bundle

Turn rough or dictated input into a reusable prompt when explicitly asked: 把这段整理成 prompt / 提炼成提示词 / 帮我写成可复用的指令 / turn this into a prompt. Preserve corrections, intent, examples, and constraints; keep the user's language. Do not intercept a task the user wants executed, and do not fire on ordinary requests that merely contain instructions.

m1nga Bundle

Wind-tunnel any instruction text that configures AI behavior — a SKILL.md, system prompt, CLAUDE.md rule, agent definition, or plugin command — by projecting it into persona × scenario simulations before it ships. Derives personas from the owner's real contexts, predicts trigger decisions from the target harness’s pre-load routing metadata, walks the body line-by-line per scenario, and reports false-fires, missed triggers, sibling-skill collisions, stale-world failures, silent-failure paths, and stranger-usability gaps, each with line-anchored fixes. Use for "probe / wind-tunnel / stress-test this skill, prompt, or instruction", "will this description misfire?", 风洞 / 场景推演 / 会不会误触, or after writing/editing any SKILL.md or long-lived prompt. NOT for experiencing a runnable product with a UI (use product-experience-officer), NOT for interrogating a plan (thinking-partner's interview mode), NOT for evaluating model outputs (write an eval grader instead), NOT for unbuilt ideas (thinking-partner).

alexandrpolansky-stack Skill

Sáhni po tomhle, když stavíš nebo styluješ konkrétní Google Sheets report, dashboard nebo přehled pro člověka (výstup automatizace, reporting, KPI přehled, sdílená tabulka). Sheets-specifické, v TrustSoft identitě, a Claude umí styling rovnou aplikovat na živý Sheet přes Sheets API bez copy-paste. Spouštěj PROAKTIVNĚ a sám (model-invoked), i když to uživatel neřekne výslovně, kdykoliv vzniká Sheet co má vypadat profesionálně. Na obecná UX rozhodnutí mimo Sheets (weby, appky, e-maily, dokumenty) použij design-advisor.

codecoradev Skill

Persistent memory engine for AI agents via the uteke CLI — remember, recall, search, forget with hybrid + fusion semantic search (fusion default since 0.16.0), documents, knowledge graph, rooms, tiered memory, and multi-agent namespaces.

3322810958-sudo Skill

Offload bounded code drafting, function decomposition, test-case drafting, bulk text extraction, classification, and summarization to the private models in D:\CodexLocalAI. Use when the user explicitly asks for the local model or when a large mechanical intermediate step can be reduced locally before Codex reviews it. Do not use for simple answers, current web research, safety-critical calculations, or final engineering decisions.

katelovescode Bundle

Use when a repo's CLAUDE.md, memory, hooks, or skill list haven't been reviewed in a while, or preferences/rules feel possibly stale, contradictory, or duplicated — audits every file that loads into a fresh session and reports what will actually reach the model.

m1nga Bundle

TaskDock keeps a multi-session task in one portable folder (TASK.json plus README, STATE and PLAN) so a later session, another agent (Claude Code, Codex, ChatGPT) or another machine can resume it, find it after a move, and reorganize its files with a reversible plan. Use when work will outlive this conversation or hand off to someone else, or when the user asks to create, resume, find, hand off, or reorganize a task folder (任务文件夹 / 续接任务 / 接手 / 找回任务 / 整理任务文件). Not for one-off questions, tiny edits, or code that already lives in its own git repository.

noskillish Skill

Answer questions about the user's own bank accounts and money using the bank MCP tools (list_accounts, get_balances, get_transactions, watches). Use whenever the user asks about balances, spending, income, subscriptions, whether a payment arrived, or wants a monthly review. Encodes the account map and categorisation rules so answers are consistent.

sofild Bundle

指导AI coding工具构建生产级Agent系统。当用户需要设计、实现或优化AI Agent系统时触发。 特别适用于：Agent架构设计、Harness工程、工具系统、权限模型、上下文管理、多智能体协作、 记忆系统、安全沙箱、MCP集成、Agent评测、长时运行等场景。支持从概念设计到生产部署的完整Agent系统构建。 触发关键词：构建agent、创建智能体、agent系统、自动化工具、AI助手、多智能体、 agent优化、升级agent、agent脚手架、agent项目模板、agent框架、 harness engineering、上下文工程/context engineering、Agent评测/agent eval、 长时运行/long-running agent、跨会话交接、agent skills构建、工具设计、成本预算控制。 核心目标：帮助AI coding工具生成结构完整、可扩展、生产就绪的Agent项目， 并证明它可靠、可控、在预算内，而非仅提供最小化demo代码。 版本演进： - v1版本：提供理论文档和最小示例，AI coding工具需要自己理解如何应用 - v2版本：提供分阶段构建指南，每个阶段都有明确的理论指导、实践步骤、检查清单和常见问题 - v3版本：引入三级构建规模、代码生成禁止复制策略、技术栈分级推荐、抽象接口引导模式 - v4版本（当前）：补齐评测(Phase 8)、确定性验证回路、预算模型、长时运行跨会话交接四大结构缺口； 修正两处范式方向——上下文压缩改为"遮蔽优先"且阈值下调到60-70%、工具系统从"多而全"反转为 "少而精+代码执行"；统一错误分类(ErrorKind)与可重试语义；同步Agent Skills开放标准、 MCP 2026(Streamable HTTP)/A2A/AG-UI；消除文档内部矛盾。 AI coding工具必须根据用户需求生成定制化代码，不得直接复制reference或example中的代码。

huggingface Bundle

Papers With Code MCP tools for searching and reading AI/ML papers, discovering recent and trending research, finding related work and paper lineage, browsing tasks, methods, conferences, organizations, and frameworks, and inspecting benchmark leaderboards with model-size and metric filters through the public Papers With Code catalog. Use whenever the user asks to find papers, survey literature, compare research, inspect an arXiv paper, explore AI/ML taxonomy or conferences, discover benchmarks or state-of-the-art models, or mentions Papers With Code, the PwC MCP server, or paperswithcode.co/mcp.

andersonadelino Bundle

Cuts silence from ONE video with auto-editor and normalizes the audio to -14 LUFS. With the `--fillers` flag, also removes filler words, stutters and duplicate takes (the speaker restarting a segment after an error), using word-level transcription plus agent judgment. Triggers "cut the silences", "remove the pauses from this video", "trim the dead air", "speed this video up by cutting pauses", "corta os silêncios", "tira as pausas do vídeo", "tira os tempos mortos", "tira os né/tá do vídeo", "corta as gagueiras". Do NOT use for creative editing (choosing takes, burned-in captions, color grading, overlays, cutting by content) — this skill only removes silence and parasitic sound. If the request is vague like "edit my video", ask what exactly before triggering.

githubstig Bundle

Write and maintain a project's docs folder (product, architecture, decision records, feature spec/plan/tasks, development notes) and an AGENTS.md step by step while the project is being built prompt by prompt, so decisions and requirements live in the repo instead of only in the conversation. Also works in small reviewed steps and makes conventional commits on request. Use only when someone explicitly asks for it, e.g. "start docs-as-you-go", "document this as we build", "resume docs-as-you-go"; never start it on your own.

abhinand5 Bundle

Run long-horizon autonomous research/agent loops with durable memory and evidence discipline. Use when the user wants to: initialize a research loop, start a new loop, generate loop artifacts (GOAL.md, ADVISORY.md, STATE.md), draft a GOAL for a phase, write an advisory/task-DAG for a loop, close out a loop, write FINDINGS from CLAIMS, spot-check loop claims, or asks about agent loop mechanics. Not for executing the loop's tasks themselves.

alexgoller Bundle

Use when the user asks for an agentic AI risk assessment, AI agent risk, shadow AI assessment, or which endpoints run Claude, ChatGPT, Cursor, Copilot, Codex, Gemini, Windsurf or similar agents on an Illumio PCE. Also use for: what are the AI agents talking to, AI agent reach, an AI agent Sankey or flow diagram of AI processes to internal systems, or assess agentic AI in the org, the PCE or this tenant. Not for general PCE health, ransomware or segmentation assessments (those are illumio-pce-assessment and illumio-daily-assessment). Requires the illumio-mcp MCP server.

aropan Skill

Use for any Django database schema or data change in CLIST — editing models in src/*/models.py, creating/running migrations, adding/renaming/removing model fields or tables, or data migrations. Ensures additive, reversible, non- destructive changes. Keywords: migration, makemigrations, migrate, schema, model field, database, ORM.

habit-hooks Skill

Spawn a reviewer sub-agent to assess a change set against habit-hooks's coding principles. Use AFTER habit-hooks reports clean — habit-hooks catches structural smells; this catches what it cannot (correctness, tests, design, missed edge cases).

habit-hooks Skill

Write or revise a habit-hooks coaching prompt. Use when a linter / knip / jscpd rule fires and the agent's default fix is wrong or shallow, or when adding a project-local override prompt. Keeps prompts short and outcome-focused using the ROSE pattern.

heznpc Skill

Record a web walkthrough with take-a-repo, including agent-authored Korean or other localized captions. Use when the user asks to record a demo, make a Korean introduction video, 한국어 소개 영상, or show a newly built web app. Outputs webm plus mp4 and thumbnail when ffmpeg exists; a walkthrough is not a functional test or voice narration.

rootsreggaeragga Skill

Doprowadza zadanie do końca bez zadawania pytań i bez promptów o uprawnienia — przy każdym wyborze bierze wariant rekomendowany, zapisuje założenie i idzie dalej, aż wszystko jest zrobione. Use when the user wants autonomous uninterrupted execution, or says "task", "sam wybierz", "bez pytania", "w pętli", "nie pytaj mnie".

almadotteam Skill

Connect an MCP client or coding agent to Alma's MCP server at mcp.alma.team and use its alma_* tools. Use when configuring Alma as an MCP server in Claude Code, Cursor, or another MCP client, or when working with alma_* tool calls.

jenkinsjamall Bundle

Actually WATCH a video (YouTube, TikTok, Instagram, Vimeo, Loom, a local file — 1,500+ sites) and write a structured note on what it teaches: the method as steps, the tools named with their real price, what was SHOWN versus merely CLAIMED, what was skipped, and a USE / TRY / SKIP verdict. Uses the claudetube MCP server (free, runs locally, no API key). Trigger whenever the user pastes a video URL, or says "watch this", "what does this video teach", "summarize this video", "what's the method in this", "is this real", "take notes on this video", "what happens at minute N", or "/video-watch". If claudetube is not installed, this skill walks the user through the one-time setup (Mac or Windows) instead of guessing at the video's contents. Never summarize a video from its title, thumbnail, or memory — watch it or say you can't.

jueane Skill

处理工具调用与本机排查的安全边界。任务涉及查原因、排查、诊断、日志检查、只读验证、MCP、带界面的 Chrome、人机验证、第三方可执行文件或本地大模型时使用。

decodingai-magazine Bundle

Build and query an LLM-maintained wiki from notes, articles and GitHub repos, with one subagent per source (modes: ingest, query).

decodingai-magazine Bundle

Build and query a minimal LLM-maintained wiki from local markdown notes (modes: ingest, query).

decodingai-magazine Bundle

Build and query an LLM-maintained wiki that learns from being used — questions, notes and repo answers land back in the graph (modes: ingest, query).

jordanbyron Bundle

Use when designing models, planning features, structuring controllers, or making architectural decisions in a Rails app. Triggers on questions like "how should I model X", "where does this logic go", "should I make a service object", adding new resources, or restructuring existing domain logic. Use this skill whenever the user is thinking about where behavior lives in their app.

mydnicq Skill

Control Herdr, a terminal multiplexer for coding agents. Use only when the user explicitly mentions Herdr or asks to use Herdr to inspect or control panes, tabs, workspaces, commands, or another agent. Do not use merely because a task could benefit from a background terminal, delegation, or parallel work. Requires HERDR_ENV=1.

mydnicq Skill

Create new agent skills or update existing ones following the Agent Skills specification. Use when the user asks to create a skill, update a skill, modify skill files, or work with SKILL.md files. Always fetches the latest specification from https://agentskills.io/specification.md before creating or updating.

niharnm Skill

Delegate a user-authorized task between Grok Bot and Codex, Claude Code, Cursor, or Antigravity through the shared Grok Bridge CLI, then verify the result and continue. Use for an explicit cross-agent handoff, not Grok Build or ordinary xAI API requests. Requires separate CLI installation and destination access.

olbartek Skill

Set up a project for multi-agent compatibility (Claude, Gemini, OpenAI Codex). Creates shared .agent directory with commands, rules, skills, and agent-specific directories (.claude, .codex, .gemini) with symlinks. Handles migration of existing agent configs.