Simulate a real-world package compromise (like the LiteLLM 1.82.7–1.82.8 incident) and practice the workflow:
detection, blast radius estimation, containment, secret rotation, and postmortem generation.
Ready
Tabletop Simulator
difficulty: mediummode: ai-proxy
You can change the preset before you click Start scenario.
Controls affect how quickly you detect compromise and how costly containment is.
Used for logs, alerts, and the postmortem. This app does not fetch anything.
Try a safe version like 1.82.6 and see your risk score change.
Footprint influences blast radius and urgency.
EDR telemetry
Improves detection timing; adds some false positives.
Network egress logs
Helps spot credential exfil domains and unexpected archives.
Version pinning
Reduces exposure; may slow patching and increase toil.
Risk score0
Estimated secrets exposed0
Time to detection—
Containment cost—
Secret Rotation Planner
Rotation completeness0%Residual risk—
Real incidents often require rotating everything accessible to the compromised process.
Edit this checklist; it exports with your report.
Postmortem Generator
This is a local text generator (no API). It uses a small rule-based template engine.
Export to JSON or copy to clipboard. The report includes citations to the original trend source.
Copied
Scenario link copied to clipboard.
How this lab works
You are the on-call responder for an AI proxy service that pulls models from multiple vendors.
A dependency in your stack is compromised.
Your job is to minimize impact while preserving evidence and restoring safe service.
Core loop
Click Start scenario.
Advance time and watch signals appear (or not).
Take response actions: isolate, rollback, rotate secrets, hunt persistence.
Generate a postmortem and export a report you can share.
Keyboard shortcuts
Space advance 15 minutes
R reset
S export report
G generate postmortem
What is simulated?
Risk scoring based on version, footprint, and controls.
Signals (EDR, outbound archive, DNS to suspicious domains) with uncertainty.
Tradeoffs: faster containment can increase downtime; deeper investigation takes time.
Secret inventory and rotation completeness.
Postmortem draft generation using local templates.
Source inspiration: The Hacker News report on the LiteLLM incident.
URL: https://thehackernews.com/2026/03/teampcp-backdoors-litellm-versions.html