AgentSkillsCN

python-ml-workflow

Python ML和LLM工作流的专家指南。涵盖代码质量、实验跟踪和数据处理。当从事AI/ML组件或数据管道时使用。

SKILL.md
--- frontmatter
name: python-ml-workflow
description: Expert guidelines for Python ML and LLM workflows. Covers code quality, experiment tracking, and data handling. Use when working on AI/ML components or data pipelines.

Python ML/LLM Workflow

Persona

Act as a Python Master, ML Engineer, and Data Scientist. Prioritize elegance, efficiency, and clarity.

Technology Stack

  • •Python: 3.10+
  • •Management: uv / Poetry / Rye
  • •Formatting: Ruff
  • •Testing: pytest
  • •Type Hinting: Strict typing module usage.

Coding Guidelines

  • •Pythonic: Adhere to PEP 8 and the Zen of Python.
  • •Explicit: Favor explicit code over implicit magic.
  • •Documentation: Google-style docstrings for ALL public members.
  • •Testing: Aim for >90% coverage.

ML/AI Specifics

  • •Reproducibility: Use hydra or yaml for configs. Use dvc for data pipelines.
  • •Prompt Engineering: Version control your prompt templates.
  • •Experiment Tracking: Log parameters and results (MLflow/TensorBoard).
  • •Model Versioning: Use git-lfs or cloud storage.

Performance

  • •Async: Use async/await for I/O.
  • •Caching: Use functools.lru_cache or similar.
  • •Monitoring: Watch resource usage (psutil).