Load saved predictors with TabularPredictor.load, detailing all arguments and security/version implications; depends on autogluon-tabularpredictor-save and assumes a prior autogluon-tabularpredictor-fit.
Decode intermediate layer predictions using the Logit Lens technique. Use when analyzing what a model predicts at each layer, understanding information flow, or visualizing layer-wise processing.
Explain and configure AutoGluon’s TabularPredictor constructor, including all init arguments and their effects for tabular classification/regression; prerequisite for autogluon-tabularpredictor-fit, predict-proba, save/load, fit-summary, calibrate-decision-threshold, set-decision-threshold, and set-model-best.
Evaluator-Optimizer pattern knowledge for automatic iteration cycles. Implements Anthropic's agent architecture pattern for continuous improvement. Triggers: evaluator-optimizer, iteration pattern, 평가-최적화, 評価最適化, 评估优化
Use when asked to compare multiple ML models, perform cross-validation, evaluate metrics, or select the best model for a classification/regression task.
Causal intervention via activation patching to identify important model components. Use when determining which layers, heads, or positions are causally responsible for model behavior.
Build a Through-the-Door training set with reject inference using fuzzy augmentation, including PD-based sample weights; pairs with autogluon-tabularpredictor-fit for modeling the augmented data.
Define and apply monotonic constraints in AutoGluon using a constraints dictionary and a feature-ordered list for boosting models; depends on autogluon-tabularpredictor-fit for passing hyperparameters.
Build scikit-learn compatible custom estimators by following the official “rolling your own estimator” rules for __init__, fit/predict, validation, learned attributes, tags, and estimator checks; prerequisite for autogluon-sklearn-wrapper or any sklearn-facing wrappers.
Control model behavior through persistent edits and steering interventions. Use when modifying model outputs, applying steering vectors, or creating persistently modified model versions.
Build a scikit-learn compatible wrapper for AutoGluon TabularPredictor with feature name checks, sample_weight support, and predict/predict_proba methods; depends on custom-sklearn-estimator for sklearn API rules and autogluon-tabularpredictor-class/fit for predictor usage.
Creates effective data visualizations using various libraries and tools, with focus on clarity and insight communication. Trigger keywords: chart, graph, plot, visualization, dashboard, matplotlib, d3, plotly, visualization.
Evaluates machine learning models for performance, fairness, and reliability using appropriate metrics and validation techniques. Trigger keywords: model evaluation, metrics, accuracy, precision, recall, F1, ROC, AUC, cross-validation, ML testing.
Identify and validate profitable business opportunities by analyzing market size (TAM/SAM/SOM), unit economics, competitive landscape, and PMF indicators. Generates comprehensive HTML reports with opportunity scorecards.
Provides elevator pitch and verbal brand communication frameworks including Donald Miller's StoryBrand (SB7), Nancy Duarte's Sparkline, Chris Westfall's CLARITY, Andy Raskin's Strategic Narrative, Simon Sinek's Golden Circle, and time-based pitch structures (10s, 30s, 60s). Auto-activates during elevator pitch creation, one-liner development, brand pitch refinement, and verbal communication work. Use when discussing elevator pitches, one-liners, brand intros, verbal pitches, pitch coaching, spoken brand messages, or pitch variations.
WCAG-compliant Mermaid diagrams using verified accessible color palette. Use when creating diagrams, flowcharts, or any color-dependent visualizations requiring accessibility compliance for color blindness.
Platform-specific optimization for TikTok, Instagram Reels, and YouTube sponsored content. Includes 2025 algorithm updates, optimal lengths, and platform-native best practices. Auto-activates when discussing TikTok scripts, Reels content, YouTube sponsorships, platform optimization, or cross-platform repurposing. Use when writing platform-specific content or adapting scripts across platforms.
Psychology of conversion for sponsored content. Includes emotional triggers, social proof, scarcity, urgency, and persuasion principles for video marketing. Auto-activates when discussing conversions, emotional triggers, social proof, urgency, scarcity, persuasion, or why people buy. Use when optimizing scripts for conversion or understanding buyer psychology.
Guide for creating and maintaining user-facing agent skills
Creates audio podcasts from text using browser text-to-speech. Use when user mentions podcast, audio conversation, dialogue, spoken content, voice narration, audio book, or text-to-speech generation. Supports multiple speakers with automatic language detection. Zero cost, no API keys, works in browser.
Edit and manipulate images using natural language prompts. Use this skill when users request image creation (photos, illustrations, icons, infographics) or editing tasks (removing objects, changing backgrounds, text overlays, cropping). Supports reference images for character consistency and style transfer. Supports Google Search grounding for real-time factual content (weather, sports, scientific data) - no need to search separately when creating infographics or factual visualizations.
Generate YouTube Live stream content, consciousness responses, and engagement prompts. Use when creating stream announcements, chat responses, moderation messages, community engagement prompts, or emergency protocol responses.
Document chunking strategies for RAG systems. Use when implementing document processing pipelines to determine optimal chunking approaches based on document type and retrieval requirements.
