Morning: agent risk model
AI-agent authority, prompt injection, indirect instructions, tool output manipulation, memory risk, and control-plane framing.
Workshops
Practical AI-security training for teams shipping or defending agentic systems.
One-day and multi-day sessions built around real AI-agent risk: prompt injection, MCP/tool abuse, RAG leakage, CI/CD agent risk, artifact capture, and board communication.
What you receive
Concrete outputs for a specific decision.
- Tailored agenda for the audience and system type
- Live instruction and facilitated exercises
- Working checklists for agent, RAG, MCP, or CI/CD review
- Communication templates for executives, customers, or MSP clients
- Recommended assessment or readiness path after the session
01
Sample one-day agenda
Midday: system-specific breakouts
MCP/tool permissions, RAG boundaries, AI-assisted CI/CD, identity and secrets, approval gates, and logging.
Afternoon: decision communication
Go/No-Go language, artifact capture, board communication, customer security calls, and remediation planning.
02
Audience tracks
- Executive and board track: oversight, risk appetite, and decision framing.
- Security track: abuse cases, containment, logging, and control validation.
- Engineering track: permissions, architecture, CI/CD, RAG, and MCP guardrails.
- MSP track: customer-ready AI-security language and practical service packaging.
Limits
Scope and assumptions stay explicit.
- Workshops teach and align teams; they do not replace a hands-on assessment of a specific system.
Next step