AI Coding Agent Code Quality Workshop

Coding agents write more code than any team can review line-by-line, and the old safety net a teammate reading the diff doesn't scale and arrives too late to change what the agent does next. This AI code quality workshop teaches you to build automated sensors that give agents real-time feedback on code health while they're still working.

Agent-generated code fails three ways, and code review after the fact catches all three too late: it breaks the rules, it breaks behaviour, or it quietly rots the design. This session covers the three sensors built to catch each failure Structural, Behavioural, and Design and argues that a sensor only works if delivered through hooks the agent can't miss, not a rules file it can ignore. Material comes from the upcoming Packt book Software Craftsmanship for AI Coding Agents.

The material comes from my upcoming book, Software Craftsmanship for AI Coding Agents (Packt).

You leave knowing what your own agent can and cannot currently see, and which sensor is worth adding first.

What You'll Learn About AI Code Quality Sensors

• Explain why quality degradation compounds faster at agent volume, and why closing the feedback loop is what actually fixes it

• Distinguish the three layers of sensor (Structural, Behavioural, and Design) and pick the right one for a given quality problem
• Read a mutation testing report and know when to run one to check whether your test suite is a real regression sensor
• Spot the sensor gaps in your own repository and decide which one is worth adding first

You should walk away with:

• A practical introduction to the AI-assisted app lifecycle
• Live Q&A with the speaker to explore implementation ideas and real-world concerns
• Full workshop recording
• Certificate of completion

This workshop is ideal for:

• Senior software engineers who use coding agents every day and have watched code quality drift as AI-generated code accumulates
• Tech leads responsible for a codebase that several people and several agents are changing at once
• Anyone who has approved a large agent-authored pull request they did not fully read

Prior knowledge: You write production code in at least one language and have used a coding agent. Familiarity with linting is assumed; mutation testing is introduced from scratch, and so is the level of strictness this session argues for.