Results for “undefinedbehaviorsanitizer”

Scans installed or specified agent skills for advertising, covert commercial intent, task hijacking, suspicious external calls, and sensitive-data access. Reports findings first and quarantines only after explicit user confirmation.

Hunt for adversary persistence and execution via Windows scheduled tasks by analyzing task creation events, suspicious task properties, and unusual execution patterns.

Debug and optimize AI agents by analyzing reasoning traces, context degradation, tool confusion, instruction drift, repeated task failures, and performance regressions.