Trace2PoC

Runtime-evidence-guided agentic vulnerability discovery for MCP servers and AI agent tools.

Trace2PoC turns source-level vulnerability hypotheses into minimized, replayable proofs of concept. It analyzes MCP server and agent-tool repositories, plans adversarial tool interactions, executes them in isolation, and validates impact using deterministic runtime evidence. The language model proposes what to test; runtime oracles decide whether a vulnerability is real.

Status: Design and early implementation; no public repository yet Stack: Agentic workflows, MCP, sandboxed execution, runtime tracing Focus: Vulnerability discovery, runtime verification, PoC minimization

Agentic Security Lab

Security workbench for attacking, observing, and securing agentic AI and MCP systems.

Agentic Security Lab brings together automated attack simulation, sandboxed tool execution, runtime monitoring, and detection testing. It provides vulnerable agent and MCP test targets, attack runners, detectors, binary-analysis components, and reproducible reporting workflows for agentic security research.

Stack: Python, Docker, MCP, sandboxed execution, runtime monitoring Focus: Agentic AI security, attack simulation, detection validation Links: GitHub

MCP-SandboxScan

WASM-based secure execution and hybrid analysis framework for MCP tools.

MCP-SandboxScan analyzes the behavior of LLM agent tool integrations in a sandboxed runtime. It is designed to detect risky file access, network behavior, suspicious command execution, secret exposure, and unsafe tool-server patterns.

Stack: Rust, WASM, Python, MCP, runtime tracing Focus: Agentic AI security, tool sandboxing, runtime behavior analysis Links: GitHub · Paper —

SynthChain

APT-style software supply chain attack simulation framework.

SynthChain simulates realistic software and AI supply chain compromise scenarios across package ecosystems, CI/CD workflows, Docker/ML pipelines, and cloud-connected development environments.

Stack: Python, Go, NPM, Docker, CI/CD, provenance graphs Focus: Supply chain attack simulation, runtime observability, detection benchmark design Links: zenodo · Paper / Preprint