AI Security Engineer Foundations + Certificate — Micah Silverman, Snyk
Source
youtube.com
Author
AI Engineer
Date
Why it matters
Makes the point that a system promptThe standing instructions a model receives before any user input — defining its role, rules, tools, and tone for the whole conversation.Full definition → does not protect data the model can reach; a restricted view does. Gives a checklist for securing agentic development: isolation, scoped credentials, human approval, verified packages.
Key takeaways · AI-distilled
In Silverman's grades chatbot exercise, the assistant's refusal itself revealed that it could reach the confidential grades. His fix limits the student assistant's data access, so the data is no longer in the model's reachable context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → at all.
A live inventory demo in Snyk's EVO catalogs the models, datasets, 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 and skills in a project; adding a policy immediately flagged a prohibited model already in use. Silverman proposes turning these AI bills of materials into CI build gates.
The MCP section separates malicious tool descriptions (tool poisoning) from ordinary code flaws, such as a command injection bug in a Figma MCP server, and recommends pairing 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 →-focused scans with dependency and source code analysis.
In the closing capture-the-flag, several audience prompts failed to extract the grades. Silverman uses that to stress that model behavior varies between attempts, so red-team tests need repeating rather than trusting a single pass.
Terms in this piece · Glossary
MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
system prompt — The standing instructions a model receives before any user input — defining its role, rules, tools, and tone for the whole conversation.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
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.
videoThe SDLC You Build Yourself — Andy Wong, Andrei Bocan & Shane Wolf, Atlassian
videoAgent Speedrun — Elizabeth Fuentes Leone & Sandhya Subramani, AWS
videoCan Your Agent Hear You Now? Building Live Voice Agents with Gemini — Thor Schaeff
videoBuilding Self-Learning Loops for Your Agent — Fuad Ali, Arize AI
