1 matches found
CVE-2026-27165
Attackers who can influence ACP "resourcelink" metadata can inject attacker-controlled prompt text into the LLM context, potentially altering model behavior...