AI Security Engineer Foundations + Certificate — Javier Garza, Snyk
Source
youtube.com
Author
AI Engineer
Date
Why it matters
Covers practical controls for AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → systems: inventory AI components, separate tool-definition scanning from code scanning, and constrain permissions with human approval.
Key takeaways · AI-distilled
Snyk's Javier Garza starts from an AI bill of materials, a scan that inventories a repository's models, datasets and agents, arguing teams cannot assess shadow AI components they do not know exist.
Defenses are layered: a database view limits what an assistant can retrieve, while restricted permissions, output validation and human approval constrain what it does next. Threat modeling maps assets and trust boundaries before deployment.
For skills and MCPThe Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.Full definition → servers, Garza separates scanning tool definitions for hidden instructions from scanning the underlying code for vulnerabilities, and argues guardrailsThe checks around a model that block bad inputs and outputs — filters, validators, and permission rules the model itself can't override.Full definition → must be explained as well as enforced because unexplained restrictions invite workarounds.
Terms in this piece · Glossary
prompt injection — An attack that hides instructions in content an AI will read — a webpage, email, or document — tricking it into following the attacker instead of the user.
MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
guardrails — The checks around a model that block bad inputs and outputs — filters, validators, and permission rules the model itself can't override.
videoHow Secrets Leak Through AI Agents (and How to Stop It) — Venice AI
videoFrom Zero to Leaderboard: Agent Evaluation — Wolfram Ravenwolf, Weights & Biases
videoBuild a Platform and Watch It Burn — Michael Forrester, Accenture & Whitney Lee, Datadog
videoContinuously Improving Agents with Langfuse — Annabell Schäfer & Lotte Verheyden, ClickHouse
