Skip to content
AgentProv

Program

Each tool checks a specific layer of the model. The catalog and the report supply the evidence.

  • AgentLock

    L2Instruction

    Authorization gate for agent tool calls that blocks consequential actions whose parameters trace back to untrusted input.

    Repositorydoi:10.5281/zenodo.21270300

  • runwarden

    L6Identity / Run

    Probe that plants a nonce in one agent run and tries to recover it from a supposedly isolated second run.

    Repositorydoi:10.5281/zenodo.22679860

  • Awesome AI Agent Attacks

    Data

    The public catalog of 385 AI agent attacks and incidents that every chart on this site is built from.

    Repository

  • 2026 AI Agent Attack Report

    All layers

    A year of agent incidents read through the eight layers: which links broke, and how often.

    Coming January

  • Provenance proxy

    Tool

    A proxy that checks provenance at the boundary between an agent and the tools it calls.

    Coming