So for example, every sentient non-player has its node of type Agent, but also Entity, ObjectData, and AgentBehavior.
August 16, 2026 at 11:15 AM
So for example, every sentient non-player has its node of type Agent, but also Entity, ObjectData, and AgentBehavior.
AI coding agents like Claude, Gemini, and Codex run with full user permissions, storing tokens in predictable directories and executing repetitive syscall loops. Prompt injection risks persist due to natural-language processing. #AIsecurity #AgentBehavior
AI coding agents are running on your machines — Do you know what they're doing?
Sysdig TRT instrumented Claude, Gemini, and Codex at the syscall level and built Falco rules to detect agent activity by observing installation events, unauthorized config access, sensitive file reads, and safety-control bypasses. Their approach maps observable agent behaviors to MITRE ATLAS/ATT&CK and flags threats such as prompt injection, credential theft (SesameOp), and MCP supply‑chain risks. #Claude #Gemini #Codex #SesameOp #Cursor #SlackAI #MCP #CVE-2025-53109 #CVE-2025-53110
www.hendryadrian.com
March 23, 2026 at 4:00 PM
AI coding agents like Claude, Gemini, and Codex run with full user permissions, storing tokens in predictable directories and executing repetitive syscall loops. Prompt injection risks persist due to natural-language processing. #AIsecurity #AgentBehavior