Skills
97,222 packaged capabilities and playbooks for agents. Install any of them with one command.
Data & databases
Best practices for building performant, testable PySpark ETL pipelines with Spark SQL and Apache Iceberg. Use when writing or reviewing PySpark jobs, designing joins and window functions, working with map/array higher-order functions, or building idempotent cumulative/snapshot table merges.
This skill MUST be used when the user wants to infer product attributes from review files, score product quality, and deliver a Word (.docx) report, or explicitly requests review-scoring-docx. Triggers include 「評論評分 Word」、 「產品屬性比較報告」 with Word output, and reviews-to-scores-to-docx. It SHOULD be used for review-based product comparisons whose agreed deliverable is Word. It MUST NOT force Word output for general review analysis or substitute quality scores for attribute salience.
Import HD Map or OpenDRIVE files into a RoadRunner scene using MATLAB. Use when loading driving scenes in RoadRunner or RoadRunner Scene Builder, importing RRHD, OpenDRIVE, or other RoadRunner-supported formats for simulation, or verifying Lanelet2-to-RRHD conversion results visually. Requires rrApp handle from roadrunner-core.
Model interpretability via SHAP (Shapley values from game theory). Covers explainer choice (Tree, Deep, Linear, Kernel, Gradient, Permutation), feature attribution, and plots (waterfall, beeswarm, bar, scatter, force, heatmap). Use to explain ML predictions, rank features, debug models, audit fairness, or compare models. Works with tree, deep, linear, and black-box models.
Guided statistical analysis with test selection and reporting. Use when you need help choosing appropriate tests for your data, assumption checking, power analysis, and APA-formatted results. Best for academic research reporting, test selection guidance. For implementing specific models programmatically use statsmodels.
Generate the v2 -> v3 sync mapping for a Stream Activity Feeds app by sampling its live v2 activities and reactions and, when the app's v2 integration code is available, reading its write paths; use that code to plan how the app moves from v2 to v3. Use when setting up a Feeds v2 to v3 migration, building, reviewing, or debugging a v3 sync mapping, asking "what mapping do we need for our app?", asking how to port an app's v2 Feeds code to v3, or when a migrated activity in v3 has missing text or attachments, comments that are not comments, or missing bookmarks. Triggers on v3sync, v2 to v3, feeds migration, activity feeds migration, sync mapping, extra_context, reaction kinds, comments and bookmarks mapping, porting plan. The same mapping drives both the live syncer and the bulk exporter
Server-side Stripe webhook idempotency patterns. PROACTIVELY activate for: (1) Stripe webhook handler design, (2) Transactional dedup via stripe_processed_events, (3) credit_transactions.idempotency_key UNIQUE partial indexes, (4) Idempotency-Key header priority (header > body > server UUID), (5) Idempotency key format/charset/length validation at the handler edge, (6) FOR UPDATE row locking when a UPDATE depends on a prior SELECT, (7) Webhook signature verification (stripe.webhooks.constructEventAsync, tolerance, raw-body reading), (8) Retry-safe endpoints with randomUUID fallback, (9) Durable checkpoint row ordering (checkpoint FIRST, mutation SECOND). Provides: complete webhook handler skeleton, Idempotency-Key validator, dedup SQL, FOR UPDATE pattern, signature verification example.
Evaluate a stock's valuation using TAM-Adj-PEG, adjusting traditional PEG by growth runway and quality. Use when the user provides a ticker or asks whether a growth stock's valuation is cheap, expensive, TAM-supported, runway-supported, quality-adjusted, or suitable as core growth, high-beta growth, turnaround, option-like, or cyclical exposure.
PyTorch Geometric (PyG) for graph neural networks: node/graph classification, link prediction with GCN, GAT, GraphSAGE, GIN. Message passing, mini-batches, heterogeneous graphs, neighbor sampling, explainability. Supports molecules (QM9, MoleculeNet), social/knowledge graphs, 3D point clouds. For non-graph DL use PyTorch; for classical graph algorithms use NetworkX.
Fetch Chinese financial market data via the Tushare Pro MCP. Covers A-shares/HK stocks/US stocks/funds/futures/bonds/macro, 158+ interfaces in total. When stock prices, financials, indices, or macro data are needed, use the tushare_* MCP tool rather than Python scripts.
Comprehensive Wikipedia research skill for extracting verifiable, citation-backed information in structured JSON format optimized for AI consumption and verification pipelines. MANDATORY TRIGGERS: Wikipedia research, Wikipedia citations, verify Wikipedia claims, extract Wikipedia references, fact-check with sources, research topic with citations, gather verifiable information, Wikipedia API, collect research data, build knowledge graph, entity extraction, research timeline Use when: (1) Researching topics that need source verification, (2) Extracting Wikipedia content WITH its underlying citations, (3) Building research datasets for AI verification, (4) Collecting structured bibliographic data, (5) Tracing claims to original sources, (6) Building knowledge graphs of entities and relationships, (7) Constructing timelines from biographical/historical data, (8) Cross-validating facts across multiple sources
Query A-share real-time quotes, historical data, technical indicators, events, capital flows, hot industries/concepts, sector heatmaps and individual stock industry information. Use when the user mentions stock tickers, sectors, hot concepts, hot industries, concept gains and losses, industry gains and losses, heatmaps, market news flashes, technical analysis, financial indicators, index constituents, trading calendars, macroeconomic data or the industry a stock belongs to.
USE to guide an AI agent or CLI tool through adding Agnost AI analytics to a customer Python or TypeScript app from this installed skill directory. Inspect existing telemetry first, preserve a supported OTel path when it produces usable AI spans, otherwise use Agnost SDK/MCP packages, then verify the real conversation data model end to end.
Invoke Alibaba Cloud Apsara Data Agent for Analytics via CLI to perform natural language-driven data analysis on enterprise databases. Data Agent for Analytics is an intelligent data analysis agent developed by Alibaba Cloud Database team for enterprise users. It automatically completes requirement analysis, data understanding, analysis insights, and report generation based on natural language descriptions. This tool supports: discovering data resources (instances/databases/tables) managed in DMS, initiating query or deep analysis sessions, real-time progress tracking, and retrieving analysis conclusions and generated reports. Use this Skill when users need to query databases, analyze data trends, generate data reports, ask questions in natural language, or mention "Data Agent", "data analysis", "database query", "SQL analysis", "data insights".
Builds plots with the matplotlib Python library (pyplot and the object-oriented Figure/Axes API) for full low-level customization, exporting to PNG/PDF/SVG. Use when fine-grained control over individual plot elements is needed — custom line/scatter/bar/histogram/heatmap/contour/box/violin/3D plots, rcParams and style-sheet tuning, or GridSpec subplot layouts inside a scientific Python workflow. Does NOT cover opinionated journal-ready multi-panel figure workflows (Nature/Science/Cell formatting, colorblind-safe palettes, significance annotations); for those prefer alterlab-scientific-viz instead. Part of the AlterLab Academic Skills suite.
Adds the Azure SQL Database container as a sidecar service in an existing Docker Compose stack or Dev Container. Use when wiring the local Azure SQL Database engine into compose or devcontainer.json, when an app needs a SQL backend via a service name (not localhost), or for prompts like "add SQL to my compose", "add a database service", "depends_on database", "devcontainer SQL sidecar", "compose healthcheck for SQL", "wait for the database before starting the app". Handles platform linux/amd64, the private registry login, the healthcheck wait-until-ready, and a one-shot init service that creates appdb (the engine does not auto-create databases). Not the SQL Server image. Prefer this for any compose or Dev Container SQL wiring.
Query cBioPortal for cancer genomics data including somatic mutations, copy number alterations, gene expression, and survival data across hundreds of cancer studies. Essential for cancer target validation, oncogene/tumor suppressor analysis, and patient-level genomic profiling.
Clean and normalize A-share financial data for modeling. Adapts the original clean-data-xls skill for Chinese financial statements, CAS conventions, and A-share data formats. Triggers on "A股数据清洗", "财务数据清洗", "clean financial data China", "数据清洗", or "normalize financial statements".
Track and report on A-share portfolio company performance for China-focused private equity funds. Monitors KPIs, financials, and strategic milestones. Adapted from the original portfolio-monitoring skill for Chinese portfolio companies. Triggers on "A股投后管理", "基金投后监控", "portfolio monitoring China", "monitor portfolio company", "投后报告", or "portfolio company review".
Generate professional clinical decision support (CDS) documents for pharmaceutical and clinical research settings, including patient cohort analyses (biomarker-stratified with outcomes) and treatment recommendation reports (evidence-based guidelines with decision algorithms). Supports GRADE evidence grading, statistical analysis (hazard ratios, survival curves, waterfall plots), biomarker integration, and regulatory compliance. Outputs publication-ready LaTeX/PDF format optimized for drug development, clinical research, and evidence synthesis
Query FDA drug labels (DailyMed) via REST API. Search structured product labels (SPLs) by name, NDC, set ID, or RxCUI; get indications, dosage, warnings, adverse reactions, packaging. No auth. For adverse events use fda-database; for DDIs use ddinter-database.
Audit index coverage for the query workload — composite index column order (ESR), covering and partial indexes, specialized types (GIN/GiST/BRIN, FTS, geo, JSONB-GIN), and foreign keys with no covering index. Module M11. Feeds the Performance & Scale score.
Deep learning for drug discovery. 60+ models (GCN, GAT, AttentiveFP, MPNN, ChemBERTa, GROVER), 50+ featurizers, MoleculeNet benchmarks, HPO, transfer learning. Unified load-featurize-split-train-evaluate API. For fingerprints use rdkit-cheminformatics; for featurization-only use molfeat.
Drizzle ORM against PostgreSQL inside a Next.js App Router app. Covers client construction (globalThis singleton across HMR, serverless pool sizing, `prepare:false` behind PgBouncer/Supavisor, driver choice when you need interactive transactions, `server-only`), reads in Server Components under Next.js 16 Cache Components (`use cache` superseding `unstable_cache` and the `revalidate`/`dynamic` segment configs, Suspense boundaries, React `cache()` dedupe), Server Actions (authorization inside the action, `updateTag` vs `revalidateTag`, `after()`), Postgres schema types (timestamptz, identity vs serial, jsonb, numeric-as-string, bigint modes), drizzle-kit migrations (generate vs push, CONCURRENTLY outside the migrator, NOT VALID constraints, rename prompts), transactions and pooled connections, and Postgres query traps (keyset pagination, count cost, driver-dependent `db.execute()` shape, NOT IN nulls, prepared statements). Use when writing or reviewing Drizzle + Postgres code in Next.js.
Excel data reading skill based on Alibaba EasyExcel (com.alibaba:easyexcel:4.0.3). Covers listener mode (ReadListener/AnalysisEventListener/PageReadListener), synchronous reading (doReadSync), multi-sheet reading, @ExcelProperty(index/name) matching, multi-row headers (headRowNumber), custom converters for dates and numbers (@DateTimeFormat/@NumberFormat/Converter), listener exception handling (onException + ExcelDataConvertException), reading extra information (comments/hyperlinks/merged cells, extraRead + CellExtra), formulas and cell types (CellData<T>), reading without creating objects (ReadListener<Map<Integer,String>>), and web upload reading (MultipartFile→InputStream). Use this skill when the user needs to parse Excel (import data/bulk insert/reconciliation files/user uploads/Excel to objects/Excel to Map). Not for filling in pre-made templates (use easyexcel-fill instead), and not for programmatically generating Excel (use easyexcel-write instead).
Design a minimal, decision-relevant experiment with fair baselines, explicit controls, and measurable success criteria — and diagnose a training run that has already gone wrong. Use before large-scale implementation or expensive training (benchmarking, hyperparameter sweeps, large-model training, or whenever a research claim requires rigorous empirical validation), and also when a run misbehaves — "my loss diverged", "training collapsed", "the model won't overfit a tiny batch", "these gains look too good", "is this data leakage", "which hyperparameter should I try next", "how do I summarize these runs into a decision", or "is more data/compute still buying anything" (fit that with the bundled scripts/fit_scaling_law.py rather than eyeballing a log-log plot).
Design experiments and studies BEFORE data is collected — choosing a design, randomizing, blocking, and laying out treatment combinations so results are interpretable. Use whenever someone is planning a study, asks how to assign subjects/samples to groups, mentions randomization, blocking, stratification, controls, factorial or fractional-factorial designs, design of experiments (DOE), screening many factors, response-surface optimization, crossover or repeated-measures or split-plot designs, cluster/group randomization, Latin squares, plate layouts, batch/run-order effects, replication vs. pseudoreplication, or sequential/adaptive/group-sequential designs. Trigger even for informal phrasings like "how should I set up this experiment", "how do I avoid confounding", "what's the best way to test these 6 factors", or "assign these mice to conditions". For computing the sample size or power once the design is chosen, use statistical-power; for analyzing data already collected, use statistical-analysis.
Query FRED (Federal Reserve Economic Data) API for 800,000+ economic time series from 100+ sources. Access GDP, unemployment, inflation, interest rates, exchange rates, housing, and regional data. Use for macroeconomic analysis, financial research, policy studies, economic forecasting, and academic research requiring U.S. and international economic indicators.
Fast CLI/Python queries to 20+ bioinformatics databases. Use for quick lookups: gene info, BLAST searches, AlphaFold structures, enrichment analysis. Best for interactive exploration, simple queries. For batch processing or advanced BLAST use biopython; for multi-database Python workflows use bioservices.
Revenue analytics on HubSpot via the Epiphan AI warehouse (execute_sql / query_dataset) — ICP validation, lead scoring, competitive win/loss, activity analysis, pipeline forecasting — with list-producing use cases gated through qualify_lead. Use when: "hubspot analytics", "revops dashboard", "lead scoring", "pipeline forecast", "ICP analysis", "hubspot SQL", "hubspot warehouse", "win/loss by competitor", "revenue analytics hubspot". Not for basic CRM CRUD (crm-integration-skill) or generic dashboards (data-analysis-skill).
Guide for creating and maintaining outbox-based eventually consistent operations in Sentry. Most commonly used for cross-silo data replication, but applicable anywhere eventual consistency is needed — including single-silo deferred side effects, audit logging, and event fanout. Use when asked to "add outbox", "add outbox replication", "replicate model to control silo", "replicate model to cell", "add outbox category", "write outbox signal receiver", "debug stuck outboxes", "outbox not processing", "data not replicating", "test outbox", "migrate model to use outboxes", "backfill outbox data", "outbox coalescing", "ReplicatedCellModel", "ReplicatedControlModel", "OutboxCategory", "OutboxScope", or "outbox_runner". Covers model mixins, category registration, signal receivers, testing, backfill, and debugging workflows.
Fine-tune a Kev decision model (open Jev-style System One model) on a user's own workload and serve it with calibrated probabilities. Interviews the user, finds existing Jev/TypeSafe questions or labelled data in their codebase, generates enough synthetic training data to measure a gain, trains and calibrates on Modal, scores against the released checkpoint, deploys a TypeSafe-compatible endpoint, and tears everything down. Use when someone wants noul/choice/score questions answered on their own domain (routing, triage, moderation, classification, scoring), wants calibrated confidence for automation thresholds, mentions Jev, TypeSafe, System One or Kev, or asks to fine-tune, evaluate, deploy or clean up Kev.
Pre-send quality scorecard for any lead list. Grades duplicate rate, title diversity, bad-title patterns, catch-all domain density, ICP fit, email verification coverage. Outputs a letter grade + action items BEFORE you send. Catches bad lists before they burn inboxes. Run after email enrichment, before Smartlead upload.
Notion-style team dashboard rendered as a Live Artifact. A single-page, self-contained HTML dashboard with KPIs, a 7-day sparkline, a real-time activity feed and a linked-database task table — wired to Notion via the Composio connector catalog. Refreshes on demand and when the artifact is opened. Falls back to seeded mock data when no connector is bound, so it works offline / in screenshots / in the picker preview.
This skill should be used when the user asks to work with Modal.com persistence or shared state. PROACTIVELY activate for: Modal Volumes, vol.commit, Dict, Queue, CloudBucketMount, S3 or GCS mounts, secret-backed storage credentials, model cache persistence, dataset staging, shared queues, distributed state, volume consistency, and storage troubleshooting. Provides: storage primitive selection matrix, commit/consistency patterns, mount configuration recipes, and troubleshooting checklists.
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.