Definition
A published security framework — the first from OWASP focused specifically on AI agent 'skills' (the plug-in extensions that let agents take actions) — that lists the ten most dangerous ways those skills can be abused or weaponised. It covers major agent platforms including Claude Code, Cursor, Codex, and VS Code.
Why it matters
As companies deploy AI agents extended with third-party skills, each skill is a potential entry point for attackers; the AST10 gives security and procurement teams a concrete checklist to evaluate risk before enabling agent capabilities.