> agent.log: initializing ops_dashboard

Agent Operations

The agentic AI security stack — every module an enterprise needs to secure autonomous systems. Research, frameworks, and tools from the field.

MCP Servers 3 connected

Model Context Protocol servers give agents tools — file access, APIs, databases. Each connection is an attack surface.

⬡

MCP Trust Boundaries

Which tools should an agent access? Least privilege for MCP.

⬡

Server-Side Request Forgery via MCP

When agents call external APIs, who validates the destination?

⬡

autoApprove Risks

The convenience setting that bypasses human-in-the-loop.

Agents 3 spawned

Autonomous agents make decisions, chain actions, and operate across trust boundaries. The threat model is fundamentally different.

⬡

Agent Threat Modeling

STRIDE doesn't cover autonomous decision chains.

⬡

Multi-Agent Security

When agents delegate to sub-agents, who inherits permissions?

⬡

Behavioral Baselines

Detecting when an agent deviates from expected behavior patterns.

Tools & Skills 3 loaded

Skills, hooks, and tool scoping define what an agent can do. This is the enforcement layer.

⬡

Secure Auth — Claude Code Skill

6-layer auth with PreToolUse hooks blocking insecure patterns.

⬡

Tool Scoping Patterns

tools vs allowedTools — principle of least privilege for agents.

⬡

Hook-Based Guardrails

PreToolUse, PostToolUse — enforcement before code hits disk.

Knowledge Base 7 entries

Research, frameworks, and writing on agentic AI security from the field.

📋

CISO's Guide to Agentic AI Risk

Threat model, controls matrix, governance checklist.

🧠

Cross-Session Memory Exfiltration

The attack vector auditors haven't found yet.

📚

Kiro Mastery — 5-Domain Study Guide

Spec-driven dev, steering, hooks, MCP, agents, skills, CLI.

💭

Why Agentic AI Changes Everything

A practitioner's take on the next 5 years of security.

Evals building

Testing and evaluating agent security posture. Red-teaming autonomous systems before they reach production.

⬡

Agent Red Team Framework

Systematic approach to finding what agents do wrong.

⬡

Prompt Injection Eval Suite

Testing agent resilience to adversarial inputs.

Observability building

Monitoring, logging, and tracing agent behavior in production. You can't secure what you can't see.

⬡

Agent Behavioral Logging

What to capture, what to alert on, what to baseline.

⬡

Token & Cost Anomaly Detection

When token usage spikes, something changed.