Season 6: Build Code Nexus Assist · Episode 6 of 8
Give the Assistant Tools
You play: Junior AI Developer
Included with Core programme and Certification prep
The situation
Support wants Code Nexus Assist to do two things: find a creator's order, and draft a ticket about it. One is a question; the other changes something in the world. Until now the assistant only produced text. With tools it can act — and every tool is a piece of authority handed to a system that reads instructions from documents, messages and web pages. The team has to decide what each tool may do, who has to say yes, what the assistant remembers between conversations, and how anyone can tell afterwards what it actually did.
What you'll learn
- AI-103 agents: roles and goals, tool contracts (name, parameters, side effects, approval), function calling, separating user data and tool output from instructions, object-level authorisation inside tools, scoped and per-user memory, per-agent tool grants, supervised multi-agent hand-offs, and traces as evidence.
- Foundry Agent Service agents, function calling with tool schemas, built-in tools and orchestrated multi-agent solutions.
Who you work with
Maya Chen
Senior AI Engineer · Your mentor
Lydia Roe
Security Engineer
Scenes
- 1.Two verbs, very different risk
- 2.Whose words are these?
- 3.A lookup and a gated ticket
- 4.Someone else's order
- 5.Reproduce the leak
- 6.Scope the memory, keep the agents in their lanes
- 7.Agents and tools check
What you leave with
You separated requests from untrusted tool content, tested simulated grants and approvals, reviewed the ownership flaw, and reproduced the fixture's shared-memory leak. The per-user test and specialist grants narrow these cases; production still needs server-side identity, argument validation, protected approvals and isolation tests.