KSKS Security Research
Home / Security architecture

Four visual sessions · cybersecurity-first

Governed Agentic AIfor cybersecurity.

A practical path from LLM foundations to secure agent design—built for cybersecurity practitioners and leaders, with visual explanations, interactive decisions and a working orchestrator story.

Launch interactive seriesView public repository
Agentic AISecurity architectureSkills + MCPPrompt injectionGuardrails
From a repeated cybersecurity task to governed agentic deliveryA four-stage flow through reusable skills, a bounded agent loop, security controls, and measurable delivery evidence.TASKRepeated workSKILLVersioned methodAGENTBounded loopEVIDENCEGoverned outcome

The four-session journey

The sequence deliberately starts with familiar chat use, adds reusable expertise, introduces controlled action, and only then asks the audience to secure and evaluate an agent. Each session can stand alone, but together they form one adoption story.

01
From chat to agentic work

Recognize when a prompt, project, skill, tool or agent is the right level of capability.

02
Securing agentic AI

Map prompt injection, tool misuse, state, identity and egress risks to concrete control layers.

03
Cyber orchestrator in action

See specialist skills, deterministic gates, evidence and ArchStudio produce governed delivery packs.

04
Build and secure your first agent

Design a bounded review agent, attack it, evaluate it and decide whether it is ready to ship.

Designed for a mixed cybersecurity audience

The material assumes security experience, not AI engineering experience. Technical detail is introduced through security decisions: which task to automate, what the agent may access, where authorization belongs, what must be measured, and when a human must remain in control.

For practitioners

  • Translate repeated expert work into a reusable skill.
  • Choose tools, MCP, retrieval and orchestration intentionally.
  • Apply schemas, policy, evidence and adversarial tests.

For leaders

  • Connect capability to quality, effort and delivery outcomes.
  • Separate current implementation from future-state claims.
  • Scale agency only when measurement supports the decision.

How the experience works

The standalone series contains 74 presentation scenes, progressive reveals, decision tabs, workflow simulations, speaker notes and source drawers. It is optimized for a live one-hour session per module and can also be explored independently in a browser.

Recommended delivery pattern
Start with the problem statement, ask the audience to make a choice before revealing the architecture, and use the simulations as discussion prompts—not as claims of production performance. Explainability remains a future clinic rather than expanding this foundation series into full XAI coverage.

Presentation and learning pages

The presentation and the technical-learning pages now share one public repository and one canonical site. The full-screen HTML runtime remains isolated under the presentations route, while the learning hub provides searchable explanations, FAQs, references, and a complete concept catalog.

Interactive URL/presentations/governed-agentic-ai/series/
Unified public sourcehttps://github.com/iamkaushiksaha/learning-hub

Ready to run the session?

Open full screen, use Space to reveal, and press N for speaker notes.

Launch presentation
Back to all topics