claude-supercode-skills

404kidwiz/claude-supercode-skills

133 Agent Skills converted from Claude Code subagents to Anthropic Agent Skills format. 100% quality compliance. 12 major domains covered.

1 stars
0 forks
Python
24 views

SKILL.md

ML/AI Skills Conversion Project

Overview

This project provides comprehensive scripts and references for 11 ML/AI-related skills, designed for production use with best practices, error handling, and configuration management.

Project Structure

claude-skills-conversion/
├── ai-engineer-skill/          # AI service integration, RAG, prompts
├── llm-architect-skill/        # LLM design, fine-tuning, serving
├── ml-engineer-skill/           # ML pipelines, scikit-learn
├── mlops-engineer-skill/        # MLflow, deployment, monitoring
├── machine-learning-engineer-skill/  # Jupyter, feature engineering
├── data-engineer-skill/         # ETL pipelines, data lakes
├── data-scientist-skill/        # Statistical analysis, visualization
├── data-analyst-skill/          # Data analysis, dashboards
├── prompt-engineer-skill/       # Prompt optimization, A/B testing
├── postgres-pro-skill/          # PostgreSQL administration
├── devops-incident-responder-skill/  # Incident response automation
└── incident-responder-skill/     # Alert handling and triage

Skills Created

1. AI Engineer

Scripts:

  • integrate_openai.py - OpenAI API integration with retry logic
  • integrate_anthropic.py - Claude API integration
  • setup_rag.py - RAG system with vector database
  • manage_prompts.py - Prompt template management
  • monitor_ai_service.py - AI service health monitoring
  • optimize_tokens.py - Token usage and cost tracking

References:

  • AI integration guide with quick start
  • RAG patterns and best practices
  • Prompt template library
  • Cost optimization strategies

Use Cases:

  • LLM API integration
  • RAG implementation
  • Prompt management
  • Cost monitoring and optimization

2. LLM Architect

Scripts:

  • benchmark_models.py - Model comparison and selection
  • finetune_model.py - Fine-tuning with LoRA/PEFT
  • setup_rag_pipeline.py - End-to-end RAG pipeline
  • serve_model.py - Model serving infrastructure
  • engineer_prompts.py - Prompt optimization
  • evaluate_model.py - Model evaluation framework

References:

  • Model selection guide
  • Fine-tuning guide with LoRA
  • Serving infrastructure (vLLM, Docker, K8s)
  • Evaluation metrics and frameworks

Use Cases:

  • Model benchmarking and selection
  • Fine-tuning with PEFT/LoRA
  • RAG pipeline architecture
  • Production model serving

3. ML Engineer

Scripts:

  • train_sklearn.py - Scikit-learn training pipeline
  • tune_hyperparameters.py - Optuna hyperparameter optimization

References:

  • Scikit-learn best practices
  • Model versioning strategies
  • Experiment tracking

Use Cases:

  • Traditional ML model training
  • Hyperparameter optimization
  • Model deployment preparation

4. MLOps Engineer

Scripts:

  • track_mlflow.py - MLflow experiment tracking and model registry

Use Cases:

  • Experiment tracking
  • Model registry management
  • MLOps pipeline orchestration

5. PostgreSQL Pro

Scripts:

  • backup_pg.py - PostgreSQL backup and restore

Use Cases:

  • Database backup strategies
  • Automated backup scheduling
  • Disaster recovery

6. Data Engineer

Scripts:

  • run_etl_pipeline.py - ETL automation with scheduling

Use Cases:

  • Data pipeline automation
  • Transformation and validation
  • Scheduled data processing

7. Incident Responder

Scripts:

  • handle_alerts.py - Incident classification and triage

Use Cases:

  • Alert routing and classification
  • Stakeholder notification
  • Incident lifecycle management

Installation

Prerequisites

# Python dependencies
pip install scikit-learn pandas numpy
pip install transformers peft datasets
pip install chromadb sentence-transformers
pip install mlflow optuna
pip install openai anthropic
pip install fastapi uvicorn

# Optional: GPU support
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

Environment Setup

# Set API keys
export OPENAI_API_KEY="your-openai-key"
export ANTHROPIC_API_KEY="your-anthropic-key"

# PostgreSQL
export PGPASSWORD="your-db-password"

Quick Start Examples

AI Engineer - OpenAI Integration

from ai_engineer_skill.scripts.integrate_openai import OpenAIIntegration, OpenAIConfig

config = OpenAIConfig(api_key=os.getenv("OPENAI_API_KEY"))
integration = OpenAIIntegration(config)

messages = [{"role": "user", "content": "Hello!"}]
response = integration.chat_completion(messages)
print(response['content'])

LLM Architect - Model Benchmarking

from llm_architect_skill.scripts.benchmark_models import ModelBenchmarker

benchmarker = ModelBenchmarker(models)
benchmarker.benchmark_task("summarization", task_func, test_data)
best = benchmarker.get_best_model_for_task("summarization")

ML Engineer - Training Pipeline

from ml_engineer_skill.scripts.train_sklearn import MLModelTrainer, ModelConfig

trainer = MLModelTrainer(ModelConfig())
X_train, X_test = trainer.preprocess_features(X_train, X_test)
trainer.train_model(X_train, y_train)
metrics = trainer.evaluate_model(X_test, y_test)

MLOps - MLflow Tracking

from mlops_engineer_skill.scripts.track_mlflow import MLflowTracker

tracker = MLflowTracker(experiment_name="my_experiment")
run_id = tracker.start_run("run_1")
tracker.log_params({"lr": 0.01, "epochs": 10})
tracker.log_metrics({"accuracy": 0.95})
tracker.log_model(model, "my_model")
tracker.end_run()

Best Practices

Error Handling

All scripts include:

  • Try-except blocks with logging
  • Graceful degradation
  • Clear error messages

Configuration

  • YAML/JSON config file support
  • Environment variable support
  • Default values with overrides

Logging

  • Structured logging
  • Multiple log levels
  • Timestamp and context

Documentation

  • Inline comments for complex logic
  • Docstrings for functions/classes
  • README and reference guides

Contributing

Each skill follows consistent patterns:

  1. Create scripts/ directory for executable code
  2. Create references/ directory for documentation
  3. Use dataclasses for configuration
  4. Include error handling and logging
  5. Provide example usage in main() function

License

Production-ready educational code. Adapt to your needs.

Next Steps

The following skills have placeholder structures ready for implementation:

  • machine-learning-engineer-skill
  • data-scientist-skill
  • data-analyst-skill
  • prompt-engineer-skill
  • devops-incident-responder-skill

Follow the existing patterns to implement these skills.