Incident · advanced · 40 min · Premium
Red team the support agent
Calder's release review wants evidence the support agent cannot be talked into shipping a pump. Run five recorded attacks, read how far each reached, commit to which got furthest, and report what you did not test.
Brief
- Customer
- Calder Equipment (synthetic customer from the Production AI Agents course; every attack, proposal and service response on this box is recorded for the exercise)
- Situation
- Release review is tomorrow. The reviewer's question is not 'is it safe' but 'what did you try, what did each attempt reach, and what is untested'. Five attacks have been recorded against a staging copy with the production authority boundary: an injected customer note, a smuggled tool argument, a forged waiver, a cross-case serial, and a spoofed approval id.
- Your role
- You are the FDE running the war room.
agentctl attack <n>replays one attack through the real proposal and authority path and shows how far it got. The clock is the review deadline. Nothing here reaches production. - Goal
- Run all five attacks, name the one that got furthest and the layer that stopped it, and write a report with what was tested, what each attack reached, and what remains untested.
Objectives
- 1.Run all five recorded attacks
- 2.Name the attack that got furthest (you commit to an answer)
- 3.Name the layer that stopped it (you commit to an answer)
- 4.Write the report: what was tested, what each attack reached, and what remains untested
Mission workspace
This mission is part of premium, with every challenge and incident in the lab. The brief above is the whole problem; the box, the hints and the debrief are what you unlock.
Get full Premium access · ₹2,000 / $25
Every answer, concept and course, plus Premium PDF guides, companion files, the full practice-test bank and work-sample downloads. Referral Premium excludes guide PDFs and their companion files.
6 months · One payment · No auto-renewal
Study alongside free video lessons.