Good Enough Agents

Understand the agent writing your code.

Two days of hands-on training for engineers using coding agents such as Claude Code, Codex, and Cursor.

We’re Lucian Ghinda and Stanislav Katkov. We teach how these tools assemble context, run commands, and make changes, so your team can decide what to give an agent and how to check its work.

  • Why does a session get worse as it gets longer?
  • What should go in a shared instruction file?
  • Which tools should an agent be allowed to use?
  • How do you check code faster than an agent can write it?

What we cover

Live demonstrations, hands-on exercises, and time to discuss how this applies to your team. Both of us teach on both days.

Day 1Inside the agent

  • The model and its context

    What a model receives on each request, why its answers vary, and what happens when context runs out. We also look at the organisations training models and how to follow their work.

  • The work loop

    Breaking a task into steps you can verify. Working with system instructions, conversation history, tool output, and retrieved information.

  • The software around the model

    The coding harness: how it dispatches work, runs tools, manages context, ends a turn, and asks for permission. What the tool handles on your behalf.

Day 2Working as a team

  • Shared instructions

    Instruction files such as CLAUDE.md, shared skills, and configuration your colleagues will inherit. When another instruction helps, and when it adds noise.

  • Tool access and the token bill

    What to expose to an agent, what to keep out, and how those decisions affect the context sent with each request and the tokens you pay for.

  • Checking the code

    Where tests, automation, and human review belong when an agent can produce more code than a person can read. Using verification to decide what is ready to move forward.

Four academic hours per day, with a 30-minute break.

Who should come?

Engineers, architects, and technical consultants who can read code and use a terminal. A group with similar experience levels works best.

The course covers using coding agents in development. Programming basics and building agent systems are outside its scope.

Who’s teaching

We both write software and teach other engineers. You’ll work with both of us throughout the course.

Lucian Ghinda

Ruby and Rails developer and product engineer, 20+ years in tech. Has been hands-on implementing LLM features at Indeed Flex, part of Indeed. Ran corporate training for more than five years with METRO, and has continued teaching solo since. His "Good Enough Testing with LLMs" workshop has been delivered repeatedly, online and on-premise. Creator of the Short Ruby Newsletter.

Stanislav Katkov

Software contractor, open-source maintainer, and startup founder. He has been working closely with LLMs for two years and training other engineers to use them. His research and day-to-day work centre around LLMs and developer tooling (CLI and MCP), and he builds his own project, Posh TUI, alongside other open-source projects.

Feedback on Lucian's workshops

Over the course of three workshops, Lucian helped us implement an entirely new testing paradigm that has already saved us countless hours of manual effort while also improving the quality of our codebase. The workshops themselves were engaging, interactive and well-structured, and the impact they've had on our products, and business, is hard to overstate.

Andrew Erlanger CTO, Code & Co

At the moment, LLMs might go unchecked and suggest tests that are either redundant or that miss important corner cases. During the workshop, Lucian introduced useful techniques to help developers and AI reason about our tests and the problem space.

José Valim Creator of Elixir

Practical questions

How is the training tailored to our team?

Allow three weeks for preparation after the discovery meeting.

We start with your stack, current tool usage, and the problems you want to work on. The training is then tailored to your team.

What do we need to arrange?

We deliver on your premises. Please provide:

  • Whiteboard and markers
  • Projector
  • Seating for all attendees and two trainers
  • Notes and pens
  • A laptop each, to follow the hands-on exercises

Check access to your chosen tools with IT before the session; blocked tool access limits the hands-on work.

Can we record the training?

The training is not recorded. Participants receive printed handbooks with the material.

Can you work with a larger or remote team?

Each group has 5–10 participants. For a larger team, we run the training again. In-person delivery is strongly preferred; tell us about your constraints in the discovery conversation. Discovery itself can be asynchronous if needed.

What support is available afterwards?

Follow-up calls and consultation, one-to-one or as a group of up to 10.

The two-day course is foundational. Deeper sessions are available on:

  • RAG and integrating LLMs into products
  • Improving team practices around LLM usage
  • Tool development for harnesses (MCP / CLI)
  • Cost management
  • Security and compliance when using AI coding tools
How is this different from tool onboarding?

We teach context management, the agent loop, tool access, cost, and verification. The focus is understanding how to work with coding agents and evaluate tools for yourself. We do not sell a coding tool.

Tell us what your team is working on.

Email us with your stack, the coding tools you use, and where you’re getting stuck. We’ll arrange a one-hour conversation with a team member and a team lead to work out whether the course fits.

contact@goodenoughagents.com