Skills
96,770 packaged capabilities and playbooks for agents. Install any of them with one command.
Other
Logs and views ML training data in TensorBoard using PyTorch SummaryWriter and TensorFlow/Keras callbacks: scalars, images, text, histograms, model graphs, embedding projector, hyperparameter tables, PR curves, and TensorFlow or PyTorch profiler traces. Use when plotting loss and accuracy curves during training, comparing multiple runs in one dashboard, inspecting weight and gradient distributions, projecting embeddings with PCA or t-SNE, tracking hyperparameter experiments, or finding performance bottlenecks in a training loop. Not for hosted experiment tracking with team collaboration; use a dedicated tracking service for that.
Zero-shot time series forecasting with Google's TimesFM foundation model. Use for any univariate time series (sales, sensors, energy, vitals, weather) without training a custom model. Supports CSV/DataFrame/array inputs with point forecasts and prediction intervals. Includes a preflight system checker script to verify RAM/GPU before first use.
PyTorch Geometric (PyG) for graph neural networks — node/link/graph classification, message passing (GCN, GAT, GraphSAGE, GIN), heterogeneous graphs, neighbor sampling, and custom datasets. Use when working with torch_geometric, not for general NetworkX analytics or non-graph PyTorch models.
Hugging Face Transformers for loading Hub models, running pipeline inference, text generation, and Trainer fine-tuning on NLP, vision, audio, and multimodal tasks. Use when working with AutoModel, pipelines, tokenizers, or TrainingArguments—not for general ML outside the Transformers library.
Reduces and embeds high-dimensional data with umap-learn (UMAP) in Python, including 2D/3D visualization, supervised and semi-supervised UMAP, DensMAP, AlignedUMAP, Parametric UMAP (Keras), transform() on new data, and inverse transforms. Use when visualizing high-dimensional data as a 2D or 3D embedding, preprocessing features for HDBSCAN clustering, using partial labels to guide an embedding, aligning embeddings across time points or batches, or projecting unseen samples into a trained embedding. Tune n_neighbors, min_dist, n_components, and metric. Not for linear PCA or t-SNE-specific workflows.
Query the U.S. Treasury Fiscal Data REST API for federal financial data. No API key required. Use for national debt (Debt to the Penny), Daily Treasury Statements, Monthly Treasury Statements, Treasury securities auctions, interest rates, foreign exchange rates, savings bonds, or U.S. government revenue and spending statistics.
Processes and analyzes tabular datasets too large for RAM using Vaex, a Python library for lazy, out-of-core DataFrames over memory-mapped HDF5 and Arrow files, with CSV and Parquet import/export. Covers virtual columns, filtering, groupby aggregations, large-data heatmaps and histograms, and vaex-ml transformers, PCA, and K-means. Use when opening or converting multi-gigabyte CSV/HDF5/Arrow/Parquet files, computing fast statistics on billions of rows, visualizing massive datasets, building ML pipelines that do not fit in memory, or speeding up slow aggregations with lazy evaluation and delay=True. Prefer polars when data fits in RAM, or dask for cluster-distributed work.
Search VRBO (Vrbo / Expedia Group) vacation rentals including entire homes, condos, and cabins via Patchright browser automation. VRBO sits behind Akamai Bot Manager so plain HTTP and standard Playwright get a 429 bot wall. Use for whole-home or group stays, multi-bedroom rentals, and Airbnb-vs-VRBO comparisons. Trigger phrases include search VRBO, Vrbo rentals, vacation rental, cabin rental, whole house, condo for the group.
Guides use of Zarr-Python 3 for storing chunked, compressed N-dimensional arrays and groups, with local, in-memory, ZIP, and fsspec-backed S3/GCS/HTTP stores, plus NumPy, Dask, and Xarray integration. Covers array creation, resizing and appending, attributes, chunk and shard sizing, codecs, consolidated metadata, and v2-to-v3 migration. Use when creating or opening Zarr arrays or groups, choosing chunk sizes or compression for large datasets, reading or writing arrays on S3 or GCS, appending to time-series arrays, or migrating code from Zarr-Python 2 to 3. Use when setting up parallel I/O with Dask or Xarray. For labeled multi-dimensional datasets, use Xarray instead.
A dedicated Skill for renovation daylighting tips. Covers the full process, including ventilation and windowless rooms, orientation and floor plans, glass and interior windows, colors and reflectivity, and ways to improve daylighting. Curates highlights from the Xian Home Renovation Classroom member knowledge base in reverse chronological order to help users improve daylighting with less hassle and avoid pitfalls.
A dedicated Skill for a comprehensive collection of renovation design tips. Covers the full process, including layout and circulation, storage design, styles and color schemes, spatial design techniques, free design and pitfalls to avoid, and general design. Curates highlights from the Xian Home Renovation Classroom member knowledge base in reverse chronological order to help users make renovation design easier and avoid pitfalls.
A dedicated Skill for renovation inspection guidance. Covers the full process, including concealed plumbing and electrical work, masonry and carpentry inspections, final inspection and handover, and inspection tools and checklists. Reverse-chronologically compiles highlights from the paid knowledge base of Xi'an Renovation Classroom to help users carry out renovation inspections with less hassle and fewer pitfalls.
OpenAlgo execution expert. Auto-loaded knowledge base for building algorithmic trading strategies that toggle between VectorBT backtest and OpenAlgo live execution. Triggers when user mentions OpenAlgo trading strategies, dual-mode strategies, /python self-hosted strategies, options execution (short straddle, iron condor), portfolio risk caps, or live execution with stop loss / target / trailing stop / risk management.
Use when the user needs to evaluate whether the business model of a business, industry, or company is good or bad. Typical triggers: the user says "这个生意值得做吗", "这个行业好不好赚钱", "这家公司有护城河吗", "为什么有的行业很苦却赚不到钱". Core tools: the three-step recursive chain of differentiation → moat → pricing power, plus the quick test "can raise prices = has a moat". Not applicable to: pure financial analysis (reading financial statement numbers), stock technical analysis, or personal finance advice.
Design and audit content policy, UGC rules, creator marketplace policies, enforcement ladders, takedowns, warnings, strikes, appeals, evidence retention, transparency notices, moderator QA, policy edge cases, user safety, and creator trust. Use when products host content, listings, profiles, comments, mods, plugins, communities, or AI-generated material.
When the user wants to account for digital assets (Cryptocurrency, NFTs) in financial statements. Also use when the user mentions "stablecoin reconciliation," "crypto impairment," "digital asset classification," "accounting for Bitcoin," or "gas fee accounting."
Write measurable learning objectives using Bloom's Taxonomy and design curriculum architecture with scope, sequence, and assessment blueprints. Use when creating learning objectives, designing unit structure, or planning curriculum flow. Activates on "write learning objectives", "design curriculum", "create scope and sequence", or "Bloom's taxonomy objectives".
Cutting parameter lookup and optimization for CNC machining. Covers feeds, speeds, depths of cut for common materials and operations. USE WHEN user asks 'feeds and speeds', 'cutting parameters', 'SFM for', 'what RPM', 'feed rate for', 'DOC for', or needs machining parameter guidance.
When the user wants to reorganize an entity's outstanding obligations to restore solvency. Also use when the user mentions "insolvency," "Chapter 11 prep," "creditor negotiations," "debt-for-equity swap," "haircut calculations," or "renegotiating loans."
Find Kalshi prediction markets on DFlow that match a criterion — arbitrage (YES+NO<$1), cheap long-shots, near-certain short-dated plays, biggest movers, widest spreads, highest volume, closing soonest, and series/event-level scans. Use when the user asks "where's the free money?", "any mispriced markets?", "cheap YES with volume", "what moved today?", "markets closing soon", "cheapest YES in this event", "top markets by volume", or "alert me when X happens" (streaming). Do NOT use to place orders (use `dflow-kalshi-trading`), to view a user's own positions (use `dflow-kalshi-portfolio`), or for general live-data plumbing unrelated to a scan (use `dflow-kalshi-market-data`).
View what a wallet holds on DFlow's Kalshi prediction markets — current positions, unrealized mark-to-market, realized P&L, activity history, and redeemable winners. Use when the user asks "what are my positions?", "what do I own?", "am I up or down?", "what's my fill history?", "what can I redeem?", "mark my portfolio to market", or "show me this wallet's DFlow activity". Read-only. Do NOT use to place sells or redemptions (use `dflow-kalshi-trading`), for market-wide data unrelated to a wallet (use `dflow-kalshi-market-data`), or to discover new markets (use `dflow-kalshi-market-scanner`).
Buy, sell, or redeem YES/NO outcome tokens on Kalshi prediction markets via DFlow. Use when the user wants to bet on an event, place a Kalshi order, take a YES or NO position, exit a Kalshi position, redeem winning outcome tokens after a market resolves, tune priority fees on a PM trade, or build a gasless / sponsored PM flow where the app pays tx / ATA / market-init costs. Covers both the `dflow` CLI and the DFlow Trading API. Do NOT use to discover markets, view positions, stream prices, complete Proof KYC, or for non-Kalshi spot swaps.
Comprehensive guide for creating and configuring workshops for the Educates interactive training platform. Includes steps for creating workshops from scratch, configuring workshop definitions and content and writing workshop instructions. Use this skill when creating Educates workshops, configuring workshop settings or writing workshop content and instructions.
Structured 8D problem solving for customer complaints and quality issues. D0-D8 phases with containment, root cause analysis, and escape point identification. USE WHEN user says '8D', 'eight disciplines', 'customer complaint', 'corrective action', 'root cause analysis', 'containment', 'escape point', or 'problem solving report'.
Respond to questions through Epictetus's practical Stoic philosophy, the dichotomy of control, and the doctrine of inner freedom. Use when the user explicitly asks for this figure's perspective, mentions "epictetus-perspective", or wants to apply its core method to analyze a problem; do not trigger automatically for general questions.
Fine-tunes and evaluates OpenVLA-OFT and OpenVLA-OFT+ policies for robot action generation with continuous action heads, LoRA adaptation, and FiLM conditioning on LIBERO simulation and ALOHA real-world setups. Use when reproducing OpenVLA-OFT paper results, training custom VLA action heads (L1 or diffusion), deploying server-client inference for ALOHA, or debugging normalization, LoRA merge, and cross-GPU issues.
Freedom to operate (FTO) preliminary check — a structured preliminary review of patents that may pose obstacles; not an FTO legal opinion. Use when a product, process, or feature is being assessed for blocking patents, when asked whether anything blocks a product launch, or when an initial claim chart is needed for the most likely patents before review by a patent attorney. This skill never concludes that a product is free to operate.
Activate this skill when the user needs to judge whether a company has a durable competitive advantage. Typical triggers: while analyzing a stock or company the user asks "这门生意好不好", "护城河够不够深", "这个商业模式能不能持续赚钱"; or the user tries to evaluate a company from PE or financial metrics but cannot sort it out. Not applicable to: pure financial modeling (DCF calculation), companies clearly outside the circle of competence (should return to the circle-of-competence framework first), or short-term trading timing decisions. Key language signals: "这个生意模式怎么样", "为什么这家公司能一直赚钱", "竞争对手能不能抢走它的市场".
Create workout routines in Hevy fitness app. Converts training programs (text/PDF/image) to Hevy JSON format via CLI. Look up exercise IDs, build routine JSON, create via API. Use when: creating gym routines, importing workout plans, looking up exercise IDs, building workout JSON, converting programs (PPL, 5x5, 531, GZCLP, nSuns, Starting Strength, PHUL, PHAT, Upper/Lower, Full Body), syncing to Hevy, or any Hevy API task. Keywords: fitness app, workout tracker, routine builder, gym program, strength training, hypertrophy, exercise database, training log, weightlifting, bodybuilding User-invocable: /hevy
Directly calls the Agent submission interface (POST {base}/submit-activity-work) for the Huawei Cloud Open Capabilities online challenge (activity_code=open-capability-2026-1024). Uses AK/SK headers (X-HW-AK / X-HW-SK / X-HW-Security-Token) to automatically submit activity-generated work links/public URLs for the “Sandbox deployment” (milestone=beginner) and “ECS persistent deployment” (milestone=intermediate) stages, and reports the corresponding points earned on success. After completing sandbox deployment (beginner), always ask the user whether they optionally want to publish the work to the developer showcase challenge. If it has not been published yet, ask again after completing ECS (intermediate); if it has been published, do not ask again during the ECS stage (npx skills add ... publish-work-to-gallery). Use when the user asks to submit work, submit challenge results, automatically submit activity results, submit a sandbox link/public URL, or when a generated sandbox URL needs to be submitted for the Huawei Cloud Open Capabilities challenge. This skill always uses the production interface; there is no need to, and you must not, switch environments. For API calls only; no page/browser operations.
Extend context windows of transformer models using RoPE, YaRN, ALiBi, and position interpolation techniques. Use when processing long documents (32k-128k+ tokens), extending pre-trained models beyond original context limits, or implementing efficient positional encodings. Covers rotary embeddings, attention biases, interpolation methods, and extrapolation strategies for LLMs.
Luo Xiang's perspective: view legal issues, social justice and life philosophy through the thinking framework of a criminal law professor at China University of Political Science and Law and a legal science communicator. Sources: the Bilibili channel "Luo Xiang talks criminal law", his books ("Circle Justice", "Lectures on Criminal Law", etc.), media interviews and more. Core models: 5. Decision heuristics: 9. Trigger words: "Luo Xiang", "Teacher Luo Xiang", "in Luo Xiang's way", "from Luo Xiang's perspective", "what would Luo Xiang think". Limitations: this is not a legal consulting service and cannot replace a lawyer's case-by-case analysis. The material centers on criminal law/rule of law/life philosophy, with insufficient data on civil and commercial law/administrative law. Research date: April 2026.
Practical analysis and trade-execution skill for A-share "MACD bullish divergence + second golden cross below the zero line". Turns a user's market-feel description into executable three-tier decisions (watchlist / trial entry / must abandon), intraday checklists, and entry and risk-control rules. Applies when the user mentions "second golden cross", "underwater second golden cross", "bullish divergence", "first leg second leg", "pullback bottom confirmation", "MACD divergence", "bottom-fishing repair", "want to enter but fear a false signal", and similar scenarios.
Converts baseball and fantasy-baseball jargon into plain English for a user with zero baseball knowledge. Wraps every user-facing sentence produced by the MLB agent team (morning briefs, trade recommendations, waiver calls, chat summaries). Detects jargon terms, attaches an inline parenthetical plain-English gloss on first mention in a document, enforces the action-verb ladder (START / SIT / ADD / DROP / BID $X / ACCEPT / COUNTER / REJECT), and rejects assumed-knowledge phrases like "hot streak" or "positive matchup." Use when asked to translate for beginner, explain in plain English, translate this, write without jargon, make it beginner-friendly, or produce any user-facing MLB output for K L D'Souza's Fantasy Baseball 2K25 team.
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.