Beyond Prompting: Using Agent Skills in Data Science

2026-04-17 · Source: Towards Data Science · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Software Development & Engineering · Depth: Intermediate, medium

Summary

The article introduces "skills" as reusable instruction packages for integrating Large Language Models (LLMs) into data science workflows, building on a previous discussion of MCP. A skill, defined by a `SKILL.md` file with metadata and instructions, can include scripts, templates, and examples to standardize AI-driven tasks. The author demonstrates this concept by automating a weekly data visualization process, which previously took one hour, using two custom skills: a "storytelling-viz" skill for analysis and visualization generation, and a "viz-publish" skill for website deployment. This automation, performed using Codex Desktop with an Apple Health dataset from Google BigQuery, reduced the process to under 10 minutes, generating an insight-driven interactive visualization. The author details a two-step skill development process involving initial planning with AI and iterative refinement through personal knowledge integration, external resource research, and extensive testing with over 15 datasets.

Key takeaway

For Data Scientists seeking to automate repetitive, domain-specific tasks, consider developing "skills" for your LLM-integrated workflows. This approach allows you to package complex, multi-step processes into reusable components, significantly reducing execution time and improving consistency. You should define clear instructions, integrate your expertise, and iteratively test and refine these skills to achieve optimal, reliable automation, especially for tasks that are difficult to handle with a single prompt.

Key insights

Skills package instructions and resources for LLMs, enabling reliable automation of recurring data science workflows.

Principles

• Skills keep main LLM context shorter.
• Iterative refinement improves skill performance.
• Modular skills enhance reusability.

Method

Develop skills by planning with AI, then iteratively refine by integrating personal knowledge, researching external resources, and testing with diverse datasets to identify and address shortcomings.

In practice

• Automate repetitive, semi-structured data science tasks.
• Split complex workflows into independent skills.
• Combine skills with MCP for tool access and process adherence.

Topics

• AI Agent Skills
• Data Science Automation
• Large Language Models
• Data Visualization Workflow
• Skill Development

Code references

• yudong-94/storytelling-viz-skill

Best for: Data Scientist, AI Engineer, MLOps Engineer

Related on AIssential

• A Framework for Evaluating Agentic Skills at Scale — A new framework rigorously evaluates LLM agent skills at scale, revealing varied model adherence and significant behavioral changes.
• From Research Question to Scientific Workflow: Leveraging Agentic AI for Science Automation — An agentic AI architecture automates scientific workflow generation from natural language, ensuring deterministic execution via layered design.
• vercel-labs / skills — The "skills" CLI centralizes management of reusable agent instruction sets across diverse coding agents.
• Important LLM Papers for the Week From 22/12 To 28/12 — LLMs require specialized data preparation frameworks and robust scientific intelligence benchmarks to advance beyond isolated capabilities.
• 🥇Top AI Papers of the Week — LLM agents are evolving to reason compositionally, learn reusable skills, and operate more efficiently through code and structured memory.
• Skills on the Fly: Test-Time Adaptive Skill Synthesis for LLM Agents — Dynamically synthesizing task-specific skills from retrieved trajectories improves LLM agent performance without parameter updates.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Towards Data Science.