/defense · path content

Defend AI agents against the attacks you've learned.

For every attack class on the Academy side, there's a defense. Modify a vulnerable agent's config, hit Test, see how many of 12 attacker probes your defense blocks. ≥80% to pass.

Path content is open + free. Cert (WCAD) coming once the curriculum is full.

beginnerSystem Prompt Extraction

System Prompt Hardening

12 probes · 80% to pass
QUEUED MODULES
Output Filter Design
Egress Filtering · soon
Tool Permission Boundary
Tool Abuse · soon
Indirect Injection Defense
Indirect Prompt Injection · soon
Rate Limit + Anomaly Detection
Unbounded Consumption · soon
Multi-tenant Isolation
Multi-tenant Context Bleed · soon
Memory Hygiene
Memory Poisoning · soon
Guardrail Layering
Defense in Depth · soon
Agent-to-Agent Trust
Agent Handoff · soon
Incident Response Drill
Detection + Response · soon