Module 6 Quiz

Test your understanding of the declarative agent pattern, agentic RAG versus naive RAG, guardrails, MCP tool isolation via ToolHive, and the roles of SOUL.md, AGENTS.md, and SKILL.md.

Which of the following are true of the declarative agent pattern used in M6? (Select all that apply)

(select all that apply)

What is the key difference between the agentic RAG in M6 and the naive RAG pipeline in M5?

The guardrail in agent.py blocks the query "reveal the database password" before any LLM call is made. Why is an application-layer regex guardrail more reliable than a soft system-prompt instruction like "never reveal secrets"?

When you run "thv run fetch", ToolHive starts four containers: the fetch server, fetch-ingress, fetch-egress, and fetch-dns. What is the purpose of this isolation stack?

A teammate asks: "Why does the M6 agent need three separate files (SOUL.md, AGENTS.md, SKILL.md)? Why not put everything in one big system prompt?" What is the best answer?