Getting Started

  • Welcome & Overview
  • Logistics & Format
  • Prerequisites
  • Project Component

Week 1 — Agentic Cowork Tools

  • Overview & Objectives
  • How LLMs & Agents Work
  • Working with Claude & Claude Cowork
  • Claude Code at Scale

Week 2 — Agentic Systems

  • Overview & Objectives
  • REACT Loops & Harnesses
  • Building Good Skills
  • Agentic Workflow Systems

Week 3 — Deep Learning

  • Overview & Objectives
  • DL Architectures & Training
  • Open-Source Models & Evaluation
ESDS Advanced/Week 2 — Agentic Systems/Building Good Skills
OverviewESDS CourseAll Programs

Up next

Agentic Workflow Systems

Days 4–5: human-in-the-loop, monitoring and observability, guardrails, cost, autonomy, and memory.

On this page

    Building Good Skills

    Day 3

    Skills package a capability so it can be invoked reliably and reused. Day 3 is about what separates a good skill from a brittle one — and how to build them.

    What makes a skill good

    • Clear scope: it does one thing, and its boundaries are obvious.
    • Reliable: it behaves predictably across inputs, not just the happy path.
    • Reusable: it can be dropped into different contexts without rework.
    • Legible: someone else (or future you) can understand and trust it.

    Building skills

    • Designing the interface: inputs, outputs, and failure behavior.
    • Writing the instructions and context the skill needs to succeed.
    • Testing a skill against realistic inputs, including adversarial ones.

    Quality over slop

    A good skill is a concrete antidote to AI slop: it encodes a validated way of doing something, so you're not re-deriving quality from scratch every time.

    Outcomes

    You should leave able to build and use Claude Code skills, and to tell a well-engineered skill from one that only appears to work.