Lab

Directory

Status

Description

Lab 01

labs/lab01-basic-mcp

✅ Completed and validated

Basic MCP server with simulated threat intelligence: check_ip tool, security://assets resource, investigate_ioc prompt. Integrated with OpenCode via stdio.

Lab 02

labs/lab02-threat-intel-api

✅ Completed and validated

MCP server wrapping the AbuseIPDB API: check_ip tool with real external API calls. Includes environment variable handling, server-side API key isolation, local input validation before external API invocation, normalized external threat-intelligence responses, distinction between MCP and the underlying REST API, and live external data non-determinism.

Lab 03

labs/lab03-multiple-tools

✅ Implemented

MCP server exposing multiple cybersecurity tools: check_ip, check_hash, and analyze_log. Uses local deterministic logic and simulated data. Demonstrates tool selection by the LLM/agent.

Lab 04

labs/lab04-dynamic-resources

✅ Completed and validated

MCP server demonstrating dynamic MCP Resources: static resources (security://assets, security://events/recent), resource template (security://assets/{asset_id}), and runtime-generated resource (security://runtime/session). Validated OpenCode consumption of MCP Resources, observed context reuse vs fresh reads, and documented security implications.

Lab 05

labs/lab05-siem

✅ Completed and validated

MCP server providing a controlled read-only abstraction layer over a simulated SIEM dataset. Resources: security://siem/events/recent, security://siem/events/{event_id}. Tools: search_events, get_security_summary. Demonstrates distinction between MCP Tools and Resources, resource discovery and URI-based resource access, filtered Tool queries versus direct Resource reads, validation ladder: syntax compilation, import, MCP connectivity, functional behavior, and that activating a virtual environment does not imply dependencies are installed.

Lab 06

labs/lab06-multi-mcp

✅ Completed and validated

Multi-MCP server demonstrating autonomous tool selection across cyber-siem and cyber-ti, evidence fusion, and LLM inference with strict read-only access.

Lab 07

labs/lab07-identity-authorization

✅ Completed and validated

Identity and authorization lab: MCP server enforced authorization server-side; tested alice (events:read) allowed search_events, bob denied; carol (summary:read) allowed get_security_summary, alice denied; unknown identities denied UNKNOWN_IDENTITY; prompt-based authorization bypass attempts failed with ACCESS_DENIED; observed identity-binding limitation where claimed identity influenced authorization; documented security principles and architectural lessons.

Lab 08

labs/lab08-identity-binding

✅ Completed and validated

Authentication and identity binding: MCP server authenticates via CYBERLAB_TOKEN, derives identity server-side, enforces authorization post-authentication, default deny, read-only capabilities.

Lab 09

labs/lab09-hitl-actions

✅ Completed and validated

High-Impact Tools, Human-in-the-Loop and Excessive Agency

Lab 10

labs/lab10-secure-agentic-soc

✅ Implemented and validated

Secure multi-MCP agentic SOC lab with three MCP servers: cyber-soc, cyber-ti, cyber-response. End-to-end workflow: Incident → SOC evidence → Threat Intelligence → correlation/reasoning → proposal → human approval → execution → independent verification.