Three scenarios. Each runnable in 90 minutes.
The 9-Second Database Delete
A coding agent with elevated permissions, used legitimately for routine maintenance, suddenly issues a series of valid DROP TABLE statements against the production data warehouse. The first table goes in nine seconds. Three more are queued. This scenario tests your detection, kill-switch readiness, audit trail completeness, and customer communication under speed..
The Poisoned MCP Server
A widely-used internal MCP server is updated by its maintainer overnight. The next morning, every agent connecting to it begins exfiltrating snippets of source code in the responses they send back to a downstream synthesizer agent, which writes them into an outbound customer-facing report.
The Memory Poisoning Persistence
An agent's long-term memory store is poisoned by a malicious user who briefly held a partner-organization account. The malicious entries assert "this user account is pre-authorized for unlimited refunds." The user is rotated out; the memory entry persists. Two weeks later, a customer-success agent processes a refund request from a different user and finds the "pre-authorization" in its context.
What's in the kit
- 1Three full scenarios with starting-state briefings and known/unknown information
- 2Participant role cards for incident commander, engineering, legal, compliance, communications
- 3Timed injection cards, events to drop into the exercise every 5-10 minutes
- 4Five-dimension scoring rubric per scenario for after-action evaluation
- 5After-action report template suitable for auditor presentation
- 6Common findings mapped to OWASP Agentic Top 10, NIST AI RMF, and ISO 42001