Agentic AI Meetup - Build Your Own Agent

Recently, I hosted a hands-on session titled “Build you own agent” during the Agentic AI Meetup on the Corda campus in Hasselt, Belgium. In this post I want to share the slidedeck and exercises of this hands-on session.

The repository contains my slides, which act as an introduction with some references. The exercises directory contains a set of exercises. Some are worked out with evaluations, others are merely ideas that you can work out yourself. Or you take all of them as input and start working on your own ideas!

Worked out exercises

Take these exercises as a starting point, and tweak them for your own needs. Either extend the evaluation set to ensure it fits your use case. Or tweak the SKILL.md.

• Code Reviewer: is a Pythonic code reviewer that is very aware of the PEP8 style guide.
• Commit Ghostwriter: writes commit messages for you that focus on the why of code changes and not on what the code changes are themselves.
• Journal Buddy: a skill to log what you are doing and review what you have done. It can prepare your standup notes.
• Ticket Prep: reads a ticket from Jira, searches through your codebase and lists the files that are likely to change. Both a skill and a subagent: if you use the subagent it offloads the reading and searching of source files to a separate context.

Ideas

Take these exercises as ideas with some extra pointers and ways of evaluation the skills.

• PR Review: Reviews the PRs that are outstanding, to list the ones that are most suited and enjoyable for you.
• Test Gap Finder: Find gaps in the test coverage (not based on code coverage, but based on semantic analysis of the code the existing tests).
• Meeting Effectuator: Turns meeting notes into action items.
• Documentation Drift Detector: Verifies whether the documentation is still in line with the code.