Data & Analytics

Runs Google Gemini's Deep Research Agent to autonomously plan, search, and synthesize cited long-form research reports.

Compares training runs across wandb, neptune, mlflow, tensorboard, or local result files by aligning them at the same step, and writes experiment reports.

Compares training runs across wandb, neptune, mlflow, or tensorboard by aligning them at the same training step instead of current-vs-final.

Teaches Claude to build interactive Plotly charts, 3D plots, and multi-panel dashboards with production-ready code.

Gives Claude a complete, code-first seaborn reference for turning pandas DataFrames into publication-ready statistical plots.

Guides Claude through the complete PyMC 5 Bayesian workflow — priors, NUTS sampling, convergence diagnostics, LOO model comparison, and prediction.

Pulls real, citable Bilibili video data — stats, metadata, tags and full danmaku — through login-free API calls so numbers are never guessed.

Scrapes BOSS直聘 (zhipin.com) job listings through Chrome CDP by keyword, city and filters, exporting plaintext-salary JSON/CSV plus summary reports.

Turns finished experiment results into an honest readout: lift, statistical and practical significance, risks, and a ship/no-ship call.

Fetches a researcher's PubMed record, classifies study types and author position, then charts the patterns and writes a strategy report.

Generates and runs reproducible Python/R analyses for medical research and outputs publication-ready tables, figures and manuscript text.

When the Bigdata.com (RavenPack) MCP only returns synthesized prose, this skill drops to the official SDK and /v1 REST endpoints for raw financials, prices, estimates and sentiment.

Co-creates a pure-markdown investment research wiki in Karpathy's LLM Wiki style by interviewing you to distill your own analysis framework into CLAUDE.md.

Pulls 20 core China A-share pharma tickers from Sina Finance's free quote API, ranks sub-sectors, and pushes a formatted daily report to Feishu.

Lets Claude Code query your Google NotebookLM notebooks for citation-backed answers and manage sources plus Studio outputs.

Enforces a discover → inspect → plan → query → validate workflow for any data question against a ktx-connected database, with a deep set of correct-SQL heuristics.

A ktx-specific skill that mines historic SQL query shards for recurring cross-table analytical intents and emits verified pattern evidence.

Maps dbt schema.yml models, sources and tests into ktx semantic-layer overlays and column notes without inventing physical columns.

A six-step framework for mapping supply-chain choke points and screening small-cap names across A-share, Hong Kong and US markets.

Turns text, URLs, code repos and Obsidian vaults into InfraNodus knowledge graphs to surface content gaps, questions and GraphRAG answers.