Data Modeling
Create data models: conceptual, logical, and physical models. Entity-relationship diagrams, dimensional modeling (Kimball), Data Vault 2.0, and activity schema patterns. Use when designing databases or data warehouses.
Change Soul
Modify the agent's core identity (soul.md). Use when: user wants to change their name, core values, language preference, or fundamental agent behavior. NOT for: changing personality/style (use change_persona), first-time setup (use onboarding), or API/config (use change_setting).
Spotify
Control Spotify playback — play, pause, skip, search, and queue tracks. Use when: user asks to play music, search for songs, control playback, or manage Spotify queue. NOT for: downloading music, non-Spotify services (Apple Music, YouTube Music), or audio file playback.
Weather
Get current weather and forecasts via Open-Meteo and wttr.in. Use when: user asks about weather, temperature, rain, wind, or forecasts for any location. NOT for: historical weather data, severe weather alerts, or detailed meteorological analysis. No API key needed.
Notion
Notion API for creating and managing pages, databases, and blocks. Use when: user asks to create a Notion page, query a database, search notes, add content to Notion, or manage Notion workspace items. NOT for: local file editing (use read_file/write_file), Obsidian vaults (use obsidian skill), or real-time collaboration.
Onboarding
First-time setup wizard that collects user name, agent personality, focus area, and language preference, then writes soul.md and persona.md. Use when: agent starts for the first time with no soul/persona configured, or user says reconfigure, setup, change my agent. NOT for: incremental edits to persona or soul (use change_persona or change_soul instead).
Summarize
Summarize or extract text from URLs, articles, PDFs, and local files. Use when: user asks to summarize a link, article, web page, or document, or asks 'what is this link about?'. NOT for: YouTube transcripts (use youtube skill), full document editing, or real-time news (use news skill).
Pdf Convert
Convert between PDF and image formats. Use when: user asks to convert PDF pages to images (PNG/JPEG) or combine images into a PDF. NOT for: text extraction (use pdf_reader), PDF editing, or document format conversion (DOCX/HTML).
Csv Analyzer
Analyze CSV and Excel files — statistics, filtering, grouping, and data previews. Use when: user asks to read, analyze, query, summarize, or explore tabular data in CSV, TSV, or Excel files. NOT for: database queries, writing new files, or non-tabular formats.
Youtube
Get YouTube video information, metadata, and transcripts/subtitles. Use when: user asks about a YouTube video, wants a video summary, needs a transcript, or asks to extract info from a YouTube URL. NOT for: downloading videos, live streams, or private/unlisted videos.
Tavily Search
>
Translator
>
Skill Creator
>
Data Versioning
Version datasets and models: DVC, LakeFS, Delta Lake for tracking data changes, reproducing experiments, and maintaining data lineage. Use when you need reproducibility or data governance.
Parse Dates Text
Parse and normalize dates, extract structured information from text, and build regex patterns. Use when cleaning messy date formats, extracting entities from text columns, or building parsing logic for unstructured data.
Image Gen
Generate images via OpenAI Images API (DALL-E 3, GPT-image). Use when: user asks to generate, create, draw, or design an image, illustration, logo, or artwork. NOT for: editing existing images, image format conversion, or screenshots.
Obsidian
Work with Obsidian vaults (plain Markdown notes). Use when: user asks to search notes, create notes, move/rename notes, or manage their Obsidian vault. NOT for: Notion pages (use notion skill), non-Markdown files, or Obsidian plugin configuration.
Code Runner
>
Model Usage
Track and display LLM token usage, costs, and model statistics. Use when: user asks about token usage, API costs, how many messages were sent, or model performance stats. NOT for: changing the model or provider (use change_setting), or viewing conversation content (use session_logs).
Change Setting
Modify pythonclaw.json configuration at runtime. Use when: user wants to set API keys, tokens, change LLM provider, adjust web port, or update any config value. NOT for: editing code, changing soul/persona files, or modifying skill files.
Pdf Writer
Create PDF documents from text, Markdown, or HTML content. Use when: user asks to create, generate, or write a PDF. Supports headings, paragraphs, bullet lists, tables, and basic formatting. NOT for: editing existing PDFs, filling forms, or adding images/charts.
Github
>
Llm Applications
Build production LLM applications using SOTA techniques: structured outputs, function calling, multi-agent systems, RAG with reranking and hybrid search, advanced prompting (CoT, ToT, ReAct), fine-tuning with DPO/ORPO, evaluation with LLM judges, and guardrails. Covers OpenAI, Anthropic, open-weight models (Llama, Mistral, Qwen), and orchestration frameworks. Use when building any LLM-powered app…
Dimensionality Reduction
Reduce feature dimensions: PCA, t-SNE, UMAP, SVD, and autoencoders. Use for visualization of high-dimensional data, noise reduction, feature compression, and as preprocessing for downstream ML models.
Summary Statistics
Compute and interpret descriptive statistics: central tendency, dispersion, shape, and confidence intervals. Goes beyond pandas .describe() with statistical interpretation and contextual meaning. Use when you need a thorough statistical summary of your data.
Clustering
Unsupervised clustering: K-Means, DBSCAN, hierarchical agglomerative, Gaussian Mixture Models, and spectral clustering. Includes optimal cluster count determination, validation metrics, and cluster profiling. Use when discovering natural groups in data.
Transfer Learning
Fine-tune SOTA pretrained models: LoRA/QLoRA/DoRA for parameter-efficient tuning, DPO/ORPO for alignment, Unsloth for 2x speedup, DINOv2/SigLIP for vision, ModernBERT/DeBERTa for NLP, and foundation model adaptation strategies. Includes multi-stage fine-tuning and continued pre-training. Use when adapting powerful pretrained models to your task.
Detect Outliers
Statistical and ML-based outlier detection: IQR method, Z-score, Modified Z-score, Isolation Forest, DBSCAN, and Local Outlier Factor. Use when investigating anomalies, cleaning data before modeling, or building anomaly detection systems.
Distribution Analysis
Analyze and fit statistical distributions: normality tests, distribution fitting, skewness and kurtosis interpretation, QQ plots, and distribution comparison. Use when assessing data normality, choosing statistical tests, or understanding data generating processes.
Tts
Text-to-speech — convert text to audio using gTTS or edge-tts. Use when: user asks to read text aloud, generate an audio file from text, or create a voiceover. NOT for: speech-to-text/transcription (use Deepgram or whisper), music generation, or audio editing.
Cohort Analysis
Perform cohort analysis on user engagement data: retention curves, feature adoption by cohort, engagement trends, and lifecycle analysis. Use when analyzing user retention, onboarding effectiveness, or product stickiness.
Model Deployment
Deploy ML models as REST APIs: FastAPI, Flask, Docker containerization, cloud deployment (AWS SageMaker, GCP Vertex AI, Azure ML), and serverless options. Use when taking a model from notebook to production.
Anomaly Detection
Detect anomalous data points and unusual patterns using statistical and ML methods: Z-score, IQR, Isolation Forest, autoencoders, and contextual anomaly detection. Use when investigating suspicious data, building monitoring systems, or identifying unusual behavior.
Pdf Protect
Add or remove password protection on PDF files. Use when: user asks to encrypt, password-protect, unlock, or decrypt a PDF. Supports owner and user passwords with permission controls. NOT for: digital signatures, certificate-based encryption, or DRM.
Customer Analytics
Customer analytics: Customer Lifetime Value (CLV), churn prediction, RFM segmentation, customer health scoring, and NPS analysis. Use when understanding customer value, predicting churn, or segmenting customers.
Change Persona
Modify the agent's personality and role (persona.md). Use when: user wants to change agent personality, communication style, focus area, or specialization. NOT for: changing user name or core identity (use change_soul), or first-time setup (use onboarding).
Ml Pipeline
Build reproducible ML pipelines: data ingestion, feature engineering, training, evaluation, and deployment as automated workflows using Airflow, Prefect, Dagster, or Kubeflow. Use when automating the ML lifecycle.
Trello
Manage Trello boards, lists, and cards via the Trello REST API. Use when: user asks to create/move/list Trello cards, check board status, manage task lists, or organise project work in Trello. NOT for: Notion databases (use notion skill), GitHub issues (use github skill), or local task lists.
Session Logs
Search and analyse your own conversation history from session files. Use when: user references older conversations, asks what was said before, or needs historical context from past sessions. NOT for: current session messages (already in context), non-session files, or external logs.
Google Workspace
>
Model Monitoring
Monitor deployed ML models: detect data drift, concept drift, prediction drift, and performance degradation. Set up alerts and automated retraining triggers. Use when maintaining models in production.
Choose Model
SOTA model selection guide: match your problem type, data size, and constraints to the best algorithm — XGBoost/LightGBM/CatBoost for tabular, TabPFN for small tabular, Transformers for text/vision, state space models for sequences, and modern AutoML. Includes 2025 benchmark results and production trade-offs. Use when deciding which algorithm to try.
Hyperparameter Tuning
Optimize model hyperparameters: grid search, random search, Bayesian optimization (Optuna), Hyperband, and early stopping strategies. Includes search space design and overfitting prevention. Use when improving model performance beyond default settings.
Demand Forecasting
Forecast product demand: incorporate seasonality, promotions, holidays, price effects, and external factors. Use for inventory planning, capacity management, or revenue projection.
Revenue Analytics
Revenue analytics: MRR/ARR tracking, unit economics (LTV:CAC), pricing analysis, churn revenue impact, and expansion revenue. Use when analyzing SaaS metrics, evaluating pricing, or building financial models.
System Random
Generate random numbers, UUIDs, passwords, or pick random items from a list. Use when: user needs a random integer/float, UUID, password, or random choice from options. NOT for: cryptographic randomness, shuffling large datasets, or seeded/reproducible randomness.
Funnel Analysis
Analyze conversion funnels: step-by-step drop-off rates, friction identification, segment comparison, and optimization recommendations. Use when analyzing user flows, checkout processes, or onboarding sequences.
Kpi Framework
Define KPI frameworks: North Star metric, input metrics, leading and lagging indicators, AARRR pirate metrics, and HEART framework. Use when setting up product analytics, defining success metrics, or aligning teams on what to measure.
Feature Selection
Select the most important features: filter methods (mutual information, chi-squared), wrapper methods (recursive feature elimination), embedded methods (LASSO, tree importance), and SHAP-based selection. Use when reducing dimensionality, improving model performance, or identifying key drivers.
Validate Data Quality
Assess data quality across six dimensions: completeness, accuracy, consistency, timeliness, uniqueness, and validity. Produces a data quality scorecard with actionable remediation steps. Use when auditing datasets, setting up quality monitoring, or establishing data contracts.
Eda Profile
Automated exploratory data analysis profiling: distributions, missing patterns, correlations, data types, and summary statistics in one pass. Use when starting analysis on a new dataset or when you need a quick comprehensive overview.
Correlation Analysis
Compute and interpret correlations: Pearson, Spearman, Kendall for numeric data, Cramér's V and point-biserial for categorical. Detect multicollinearity and identify meaningful feature relationships. Use when exploring feature relationships or checking for multicollinearity before modeling.
Infrastructure Design
Design ML infrastructure: feature stores, model registries, experiment platforms, and compute management. Covers architecture patterns for small teams to enterprise scale. Use when planning ML platform architecture.
Model Evaluation
Evaluate ML model performance: classification metrics (accuracy, AUC, F1), regression metrics (RMSE, MAE, R²), cross-validation strategies, learning curves, confusion matrices, calibration, and fairness metrics. Use when assessing model quality or comparing models.
Segment Analysis
Discover natural data segments and analyze group differences: clustering-based segmentation, cross-tabulation, statistical comparison between groups, and segment profiling. Use when looking for natural groupings or comparing subpopulations.
Clean Dataset
Detect and fix common data quality issues: missing values, duplicates, type mismatches, encoding problems, and inconsistent formats. Use when starting a new analysis, preparing data for modeling, or auditing a dataset for issues.
Feature Engineering
Create predictive features from raw data: encoding categorical variables, binning, polynomial features, interaction terms, date/time features, text features, and target encoding. Use when preparing data for machine learning models or enriching datasets for analysis.
System Time
Get current date, time, timezone, Unix timestamp, or convert between timezones. Use when: user asks what time it is, what date today, timezone conversion, or Unix timestamp. NOT for: scheduling/calendar logic, cron syntax, or historical date arithmetic.
Nlp Pipeline
Build NLP pipelines: text preprocessing, tokenization, embeddings (Word2Vec, BERT, sentence-transformers), sentiment analysis, named entity recognition, topic modeling, and text classification. Use when working with text data for analysis or model building.
Market Basket Analysis
Association rule mining: find products frequently bought together using Apriori, FP-Growth. Calculate support, confidence, and lift for cross-sell and upsell recommendations. Use when designing product bundles or recommendation systems.
Generate Synthetic Data
Create realistic synthetic datasets for testing, prototyping, and demonstration. Supports tabular, time series, and relational data with configurable distributions, correlations, and anomalies. Use when you need test data, want to prototype an analysis, or need to share data without privacy concerns.
Ab Test Analysis
Analyze A/B test results with statistical significance, sample size validation, confidence intervals, and ship/extend/stop recommendations. Use when evaluating experiment results or deciding whether to ship a variant.
Data Profiling Report
Generate a comprehensive, shareable data profiling report covering structure, quality, statistics, and recommendations. Designed for stakeholder communication and data documentation. Use when onboarding to a new dataset or creating data documentation.
Pdf Split
Split a PDF into separate files by page ranges. Use when: user asks to split, extract pages from, or break apart a PDF. Supports extracting single pages, ranges, or every-N-pages splitting. NOT for: merging PDFs, reading text, or converting formats.
Finance
Fetch stock quotes, crypto prices, forex rates, and financial data via Yahoo Finance. Use when: user asks about stock price, market data, company ticker, cryptocurrency price, or forex exchange rates. NOT for: options/futures data, fundamental analysis, or real-time tick-by-tick data.
Handle Missing Data
Advanced missing data imputation strategies: MICE, KNN imputation, regression imputation, domain-aware methods, and missingness mechanism analysis. Use when simple fill strategies aren't sufficient or when missing data patterns need careful treatment for modeling.
Transform Data
Reshape, pivot, melt, merge, aggregate, and transform tabular data using pandas, polars, or SQL. Use when restructuring datasets, combining multiple tables, creating aggregations, or converting between wide and long formats.
Model Interpretation
Interpret ML model predictions: SHAP values, permutation importance, partial dependence plots, individual conditional expectation, LIME, and counterfactual explanations. Use when explaining model behavior to stakeholders, debugging model decisions, or meeting regulatory requirements.
Neural Network Design
Design state-of-the-art neural network architectures: TabNet/TabPFN for tabular, Vision Transformers/ConvNeXt V2 for images, Mamba/RWKV state space models for sequences, Mixture of Experts, diffusion models, and multimodal architectures. Includes Flash Attention, LoRA, quantization, and modern training recipes. Use when designing a neural network for any modality.
Trend Analysis
Detect and analyze time-based trends: seasonality, changepoints, decomposition, and growth rates. Use when analyzing metrics over time, detecting structural changes, or understanding periodic patterns.
Slack
Slack API integration — send/read messages, react to messages, pin items, and fetch member info. Use when: user wants to send a Slack message, check Slack channels, react to a message, or manage pins. NOT for: Discord (use Discord channel), email (use email skill), or real-time Slack event streaming.
Training Optimization
Optimize deep learning training: learning rate schedules (cosine, warm-up, one-cycle), regularization (dropout, weight decay, label smoothing), early stopping, mixed precision, gradient accumulation, and distributed training. Use when improving training speed, stability, or model performance.
>
Recommendation Systems
Industrial recommendation systems: recall (DSSM, SASRec, MIND, Swing, ItemCF), pre-ranking (COLD), ranking (DIN, DIEN, BST, DCN V2, DeepFM), re-ranking (PLE, MMOE, DPP), multi-task learning, feature engineering for CTR, and real-time serving architecture. Covers the full recall→pre-rank→rank→re-rank pipeline used at top tech companies. Use when building recommendation, search ranking, or ad syste…
Http Request
>
Dbt Models
Design dbt (data build tool) models: project structure, staging/intermediate/marts layers, source freshness tests, custom schema tests, and documentation. Use when building analytics engineering pipelines with dbt.
Computer Vision
Build computer vision models: image classification, object detection, segmentation, data augmentation strategies, and evaluation. Covers CNN architectures, transfer learning from pretrained models, and handling small datasets. Use when working with image data.
Classification
Build classification models: logistic regression, decision trees, random forest, XGBoost, LightGBM, SVM, and KNN. Includes class imbalance handling, threshold tuning, and multi-class strategies. Use when predicting categories from features.
Pdf Merge
Merge multiple PDF files into a single PDF. Use when: user asks to combine, join, or concatenate PDFs. Supports page selection and ordering. NOT for: splitting PDFs, editing content, or converting formats.
Embeddings Vectors
Work with embeddings and vector databases: choose embedding models, build vector indexes, perform similarity search, and design retrieval systems. Covers FAISS, ChromaDB, Pinecone, Weaviate, and Qdrant. Use when building semantic search, recommendations, or RAG systems.
News
Search and summarize news on any topic via web_search or DuckDuckGo. Use when: user asks about recent news, current events, headlines, or wants a news briefing on any subject. NOT for: historical news archives, paywalled articles, or real-time breaking news alerts.
Regression Ml
Build ML regression models: gradient boosting, random forest, elastic net, SVR, and KNN regression. Includes model comparison, residual analysis, and prediction intervals. Use when predicting continuous outcomes with complex non-linear relationships.
Pdf Reader
Extract text content from PDF files. Use when: user asks to read, extract, or analyze content from a PDF document. Supports multi-page extraction, page ranges, and metadata. NOT for: scanned/image PDFs (OCR), PDF editing, or creating PDFs.
Web Scraper
>
Ci Cd Ml
Set up CI/CD for ML projects: automated testing (data validation, model quality gates), model registry integration, staged deployment (shadow, canary, blue-green), and rollback strategies. Use when establishing MLOps practices.
Experiment Tracking
Track ML experiments with MLflow, Weights & Biases, or DVC: log parameters, metrics, artifacts, and model versions. Compare runs, reproduce results, and collaborate with your team. Use when running experiments and needing to track what works.