Forward Deployed

Agentic Systems: sections and modules

T3.5

Robust agents: retries, fallbacks and guards

Make tool calls safe and recoverable.

About 2 h 59 min: 44 min reading + 2 h 15 min practice · 6 lessons

Scope: Single-agent reliability; platform security in T4.4
Prerequisites: Tool design, the agent-computer interface, MCP and A2A

Lessons

6 lessons

1
The Confirmation Gate and Memoization
9 min
2
Retries and Fallbacks
7 min
3
Parallel and Sequential Tool Calls
6 min
4
Observability for Tool Calls
6 min
5
The Eight Guard Checks
8 min
6
Practise Robust Agent Interview Questions
8 min

Also used in this module

Lessons taught in another module that this one builds on.

Why Agents Need State
7 min
Checkpoints, Interrupts, and Human Approval
8 min
Retries, Fan-Out, and Streaming
8 min

Practise and Check

Deliverable: P3 Debug a failing agent
Assessment: P3 rubric
Interview probe: What stops the agent doing damage?
Common mistakes: Retrying non-idempotent actions

Your Progress

How confident are you with this module?

1 means "I could not explain it yet"; 5 means "I could teach it". Rate again any time.

Your notes on this module

Saved in this browser when you leave the box.