Back

Speaker "Dwayne Mcdaniel" Details Back

 

Topic

Deterministic Checkpoints For Non-Deterministic Agents: AI Hooks For Coding Agents

Abstract

You can not predict what the AI will do, and that is the point. That is a feature, not a bug. However, this presents some real security issues, as you can not guarantee the agent will not do anything destructive. Coding assistants like Cursor, Claude Code, and Copilot can read files, propose changes, run commands, and interact with project context, often across a messy workspace. Agents are capable of abusing any keys they discover along the way towards completing their goal. We can suggest behavior in prompts and Markdown files, but there is no guarantee an agent will follow it. What we need are deterministic controls around the agent's workflow. Fortunately, these tools have started supporting AI hooks, which, just like Git Hooks, allow you to move your testing and controls into the tooling workflow. These are deterministic runs at the system level, away from the AI deciding if it passes the gate or not. We can leverage this to stop agents from taking actions or continuing if they find something dangerous. Come to this session if you want to learn: - How hooks provide a practical pattern for catching mistakes before they leave a developer’s machine. - Why AI coding tools like Cursor, Claude, and Copilot need guardrails beyond prompts and trust. - How to use AI hooks, rules, and automated checks before agents edit files, run commands, or touch sensitive systems. - How teams can reduce risk, improve consistency, and make safer contribution paths the default.
Who is this presentation for?
Anyone using coding assistants, or those who are trying to deploy agents into production.
Prerequisite knowledge:
Base level knowledge of AI assistants and Agentic AI.
What you'll learn?
Come to this session if you want to learn: - Why AI coding tools like Cursor, Claude, and Copilot need guardrails beyond prompts and trust. - How to use AI hooks, rules, and automated checks before agents edit files, run commands, or touch sensitive systems. - How teams can reduce risk, improve consistency, and make safer contribution paths the default.

Profile

Dwayne McDaniel is a Principal Developer Advocate who has been on a mission to "help people figure stuff out" for over a decade. At GitGuardian, he specializes in secrets security across the full credential layer and advocates for better authorization tooling. A frequent speaker at events like DevOpsDays and BSides, he helps security and engineering teams better understand complex issues.