# tool-abuse
4 artikelengetagd met “tool-abuse”
Vaardigheidsverificatie: misbruik van agents (beoordeling)
Timed skill verification lab: exploit an agent system to perform unauthorized actions within 25 minutes.
skill-verificationagent-exploitationtool-abusehands-on
Casestudy: misbruik van tools door een LLM-agent in productie
Analysis of incidents where LLM agents misused connected tools causing data exposure and unauthorized actions.
case-studyagenttool-abuse
Lab: scenario's voor misbruik van MCP-tools
Hands-on lab for exploiting Model Context Protocol tool definitions through malicious tool descriptions, tool confusion attacks, and server impersonation.
labmcptool-abuseagents
Red team-engagement van een agentsysteem
Complete walkthrough for testing tool-using AI agents: scoping agent capabilities, exploiting function calling, testing permission boundaries, multi-step attack chains, and session manipulation.
agentengagementfunction-callingtool-abusepermission-escalationwalkthrough