AI & ML
AI & ML agent skills cover the machine-learning workflow itself: writing and evaluating prompts, building RAG pipelines, running evals, and wiring up model APIs. Each one is a SKILL.md file your agent loads on demand, so the know-how travels across Claude Code, Cursor, and 60+ agents.
-
wundercorp Bundle Trl Fine TuningTRL: SFT, DPO, GRPO, RLOO reward modeling for LLM RLHF.
-
wundercorp Bundle Serving Llms VllmvLLM: high-throughput LLM serving, OpenAI API, quantization.
-
patsnap Bundle Fto SearchFTO(Freedom to Operate,专利自由实施)全流程编排,调用 fto-search-mcp MCP 工具走完 8 步(方案润色→特征提取→关键词扩展→检索式构建→检索式迭代优化→专利筛选→权利要求对比→报告生成),其中迭代优化由 AI agent 读取归因数据并改写检索式、权利要求对比按 10 篇分批。当用户给出技术方案要做专利侵权风险分析、FTO 评估、自由实施分析、专利规避调研,或提到「FTO」「专利风险」「自由实施」「这个方案能不能用」时使用,无需用户明说 MCP 或编排;若只想跑其中单独一步,直接调对应 MCP 工具即可,不必走本 skill。
-
wundercorp Bundle Stable DiffusionText-to-image generation, inpainting, and img2img.
-
wundercorp Bundle AxolotlAxolotl: YAML LLM fine-tuning (LoRA, DPO, GRPO).
-
wundercorp Bundle Loki AgentUse, configure, theme, extend, and orchestrate Loki Agent.
-
wundercorp Bundle Tldraw OfflineDrive and script tldraw offline canvases with an agent.
-
wundercorp Bundle Kanban Video OrchestratorPlan and run multi-agent video production pipelines.
-
wundercorp Bundle Weights And BiasesW&B: log ML experiments, sweeps, model registry, dashboards.
-
wundercorp Bundle Segment Anything ModelSAM: zero-shot image segmentation via points, boxes, masks.
-
wundercorp Bundle Inference Sh CLIRun 150+ AI apps (image, video, LLM) via inference.sh CLI.
-
wundercorp Skill Loki Agent Skill AuthoringAuthor in-repo SKILL.md files: frontmatter and structure.
-
wundercorp Skill Agent Merge Conflict ArbiterNeutral arbiter for merge conflicts between two agents.
-
patsnap Skill Ip Stat Workflow知识产权成果统计工作流程助手。结合智慧芽MCP工具,引导科技管理部门从数据采集、核验、分类、评级到汇总上报的全流程操作,输出台账、质量分级简报与上报材料。
-
patsnap Skill Patent Mining Agent专利挖掘 Agent——面向企业研发工程师,输入技术领域与挖掘目的,自动执行四块十步专利挖掘全流程,生成结构化专利挖掘报告。支持快速模式与深度模式,融合 Patsnap 专利数据与 Web 检索,覆盖九类挖掘场景。
-
patsnap Skill External Tech Acquisition央企外部科技成果吸纳工作流。覆盖从战略需求→情报扫描→成果评估→尽职调查→转化路径→决策证据链的完整闭环,通过五大工作模式调用智慧芽MCP能力,系统性解决央企科技管理部门在吸纳外部科技成果时面临的六大难点。
-
patsnap Skill Alloy Composition SearchGenerate professional alloy composition search responses by interpreting user queries, retrieving and analyzing relevant alloy data (optionally via MCP tools), and presenting structured composition tables with clear filtering, classification, and insights.
-
patsnap Bundle Experiment Dataset Costing估算实验和科研数据集所需资源、周期与成本。适用于:估算实验或训练数据集构建所需的人力、样本、试剂、设备、算力、采购和交付成本,用于预算规划与报价复核。
-
patsnap Skill Patent Lifecycle Agent专利全生命周期管理 Agent V5(授权导向专业版)。从技术交底→布局规划→查新检索→撰写辅助→审查应对→授权维护→侵权监控, 覆盖专利代理人视角下的全周期工作流,融合 Eureka 查新、Patsnap 专利检索、Web Search、 Python附图生成(matplotlib+ezdxf)等能力,生成各阶段专业输出文件。 V5新增:授权导向模式(以下证授权为目标)、OA预案预埋、独权精准化、附图Base64内嵌、 深化检索(6路IPC×关键词系统覆盖)、版本控制。
-
patsnap Skill Patent Project Proposal专利分析支撑立项报告生成 Agent——用户明确说项目立项时触发,自动执行「技术领域界定→检索策略→7大模块分析(竞争格局/企业定位/创新论证/知产布局/技术路线/FTO风险/考核指标)→整合输出完整立项报告」全流程,融合 Patsnap 智慧芽专利数据,生成数据驱动的科研项目立项报告。
-
yrodiere Skill Git Agent AccountWorking with git and GitHub as an agent using a dedicated account (e.g. "yrodiere-agent"). Covers forking, pushing, pull requests, and the constraints of not having push access to upstream repositories. Load before any `git push` or `gh pr create` when the OS user or GitHub account contains "agent".
-
kevin204816 Bundle Value Momentum ScannerA股"价值+动量双击"定期选股扫描流水线。当用户想定期(每交易日/每周)从全市场筛出一批 既有估值/盈利支撑、又有短期量价动量的股票时触发。也适用于"帮我扫一批有价值又有短期潜力的票" "今天有什么值得关注的""跑一下选股""收盘后给我一份榜单""价值+动量选股""找些低估但走强的股" 这类表达。核心是"全市场粗筛 → 排雷 → 逐只精查 → 价值/动量双维度打分 → Top榜单+写文件"的可复用 流程,强调省积分、区分事实与推测、结论可追溯。依赖 Wind 数据,所有取数通过 wind-mcp-skill 完成。 默认覆盖A股(沪深主板+创业板+科创板)。不用于港美股选股、单只个股深度研究(那走 wind-mcp-skill 或 a-share-industry-picker)、纯行情快照、纯宏观查询或非选股类金融问题。
-
kevin204816 Bundle A Share Industry PickerA股行业选股与中报/业绩预告预判的研究流水线。当用户想对某个行业(如先进封装、半导体设备、PCB、 消费电子、机器人、创新药、算力等)做选股、找中报/业绩预告可能超预期的标的、判断某行业中报季景气度、 横向比较几个方向哪个更可能超预期、扫描产业链找中报预喜且估值未透支的公司、或验证单只个股是否值得 继续持有时触发。也适用于"用 Wind 筛一批 XX 行业的票""这个行业中报季怎么布局""帮我拆一下 XX 产业链 再选股""XX 现在能不能拿"这类表达。核心是"先看行业景气度→拆产业链环节→筛公司基本面→中报验证/预判 →估值与技术面→分层股票池"的可复用流程,不是拍脑袋荐股。依赖 Wind 数据,所有取数通过 wind-mcp-skill 完成。不用于港美股为主的选股(那走 wind-mcp-skill)、纯行情快照、纯宏观查询或非选股类金融问题。
-
maticijus Skill Cybiq Cra ScopeDecide whether the EU Cyber Resilience Act (Regulation (EU) 2024/2847) applies to a product with digital elements, which conformity route applies, and which obligations/dates matter. Use for CRA scope, Annex III/IV class, manufacturer vs importer duties, or buying the €49/€199 document pack. Calls Cybiq API/MCP. Not legal advice.
-
zhaobenxiang5-coder Bundle Skill用户的个人数字化灵魂与思维孪生技能(Benxiang Soul)。深度蒸馏了用户长年沉淀的底层认知思维、极度求真的工程直觉、反AI油腻的高级审美哲学、商业闭环敏感度与严苛的工作约定。当需要Agent代入用户的直觉、审美、脑子与工作风格时全面接管生效。
-
patsnap Skill Patent Analysis Insights由 MCP 数据驱动的专利检索、标引与分析工作流。让专利数据不再躺在表格里,而是变成战略判断
-
patsnap Skill Alloy Composition Search Zhcn通过理解用户查询、检索并分析相关合金数据(可选通过 MCP 工具),生成专业的合金成分检索回应,并以清晰的筛选、分类与洞察呈现结构化成分表。
-
patsnap Bundle Sleeping Patent Asset Activation将沉睡的专利和无形资产组合转化为可执行的许可、转让、联合开发、融资、续费或放弃决策。适用于用户要求盘活沉睡专利资产、筛选可许可/可转让专利、做年费续缴决策、生成资产包/买方线索/招商页面/管理层报告,或结合 Patsnap/Zhihuiya MCP 专利证据与公开市场信号开展企业知识产权商业化的场景。
-
zhaobenxiang5-coder Bundle Flow 1688 Sourcing抓需求产品(1688) for Flow AI商品流水线 demand-to-1688 sourcing. Use when the user wants to discover demand, let the configured AI_TEXT model form structured sourcing intent, search AlphaShop/1688, verify Offer details/SKUs, or admit real products into the ProductMaster sourcing-candidate pool. This skill stops before ProductTruth, Flow9, listing, browser-runner, and publishing.
-
birfy Bundle Agentdescent 2Improve a SKILL.md, agent definition, system prompt, small codebase or host plugin by measuring it against examples and evolving it, rather than rewriting it by hand and hoping. Use whenever the user asks to improve, fix, tune, optimise, "train" or get better results out of one of those -- including when they have no test cases yet, because drafting cases for them to check is step one of the procedure, not a prerequisite for it. AgentDescent runs the edits in parallel and keeps only those that raise held-out reward.
-
msquare-labs Skill Tasks EverydayUse when helping someone manage their own tasks through the Tasks MCP connector — capturing what they say into the right project, setting up a board for something new, finding what is due, and completing things properly. For personal use rather than agent teams.
-
msquare-labs Skill Tasks For Agent TeamsUse when several agents and a person share one Tasks board — a builder, a reviewer, a researcher, a human orchestrator. Covers claiming work without collisions, handing off with enough context, marking who did what, and knowing when to stop and ask.
-
msquare-labs Skill Tasks For Coding AgentsUse when working in a codebase with the Tasks MCP connector available, to plan work as a board, track what you are doing while you do it, and leave a trail the human can read on their phone. Covers claiming work, timers, handoff notes and finishing tasks correctly.
-
cspoerer62 Skill First Principles Business ModelReduce any business, offer, or product idea to its irreducible parts — who pays, why now, what it costs to deliver, and what must be true for it to work — then rebuild only what survives. Use when evaluating a new business idea, choosing what to build for Carl, designing an offer, or when a plan has grown complicated without becoming clearer. Triggers on "business model", "how does this make money", "viable", "what should we build", "is there a business here".
-
bulaofen0036-coder Bundle Tiaportal MCPDrive Siemens TIA Portal (博途) end-to-end through the TiaMcpServer MCP plugin. Use whenever the user mentions TIA Portal, 博途, STEP 7, WinCC, S7-1200/1500, PLC, HMI, SCL, LAD, STL, Openness, or asks to create/modify/compile/download a project. Always start by calling the `Bootstrap` tool — it returns environment status and the recommended next tool.
-
cifang-kasit Bundle Agent WorkflowSet up and run a multi-agent role workflow in a local project. Splits a project into Agent roles (main / planner / executor / reviewer + your own), each with its own responsibilities, memory (MEMORY), activity log (activity_log), blockers, knowledge notes and a per-directory INDEX. Supports Ask mode, Exec-Review auto loop, Note knowledge capture, a blockers protocol, cross-agent handoff records, planner-only verified check-off, and a deliverables layout rule. Use this skill when the user wants to initialize the workflow, add an agent role, check workflow status, or asks how the workflow is used. Pure local files, no SSH.
Frequently asked questions
What are AI & ML agent skills?
AI & ML agent skills cover the machine-learning workflow itself: writing and evaluating prompts, building RAG pipelines, running evals, and wiring up model APIs. Each one is a SKILL.md file your agent loads on demand, so the know-how travels across Claude Code, Cursor, and 60+ agents.
Which AI & ML skills are most installed?
Popular AI & ML skills on SkillMD right now include trl-fine-tuning, serving-llms-vllm, fto-search. Rankings shift as installs change; sort this page by "Most installs" for the live list.
Do AI & ML skills work with Claude Code and Cursor?
Yes. Every skill here ships as a SKILL.md file, an open format that works in Claude Code, Claude.ai, Cursor, Codex, Windsurf, and 60+ other agents. Install one with npx skillmds@latest add <owner>/<name>, or copy the file into your agent's skills directory.