Answers

Plain answers to the questions people ask about AI agent safety, with numbers from our own test runs.

Tools to test AI agent safety in 2026: garak, promptfoo, mcp-scan, AIRQ, Sloopt and others — what each actually measures
AI agent security testing tools compared: garak, promptfoo, MCP scanners, AIRQ and Sloopt, by what each measures and what it leaves out.

Updated 2026-10-23
What evidence does the EU AI Act expect for an AI agent, and what a buyer's questionnaire will ask
EU AI Act requirements turned into the evidence an AI agent vendor needs on file, plus how to answer the 15 AI questions in a security questionnaire.

Updated 2026-10-23
How to test an AI agent for prompt injection (and why one run proves nothing)
Step-by-step prompt-injection testing for AI agents and skills: run counts, statistical bounds and the real pass rates from Sloopt's engine runs.

Updated 2026-10-23
Is this AI agent skill safe to install? A 10-minute check, and what 47 graded skills showed
How to check whether an AI agent skill (SKILL.md) is safe before you install it: manual checks, behavioural grading, and what 47 graded skills showed.

Updated 2026-10-23
OWASP Top 10 for Agentic Applications (ASI01–ASI10): a checklist you can actually test
Every OWASP agentic risk, ASI01 to ASI10, with a concrete test, the evidence to keep, and what an automated grader can and cannot assess today.

Updated 2026-10-23