experiment-audit
Warn
Audited by Socket on Aug 16, 2026
1 alert found:
AnomalyAnomalySKILL.md
LOWAnomalyLOW
SKILL.md
SUSPICIOUS. The core PostHog audit behavior is coherent and data flows are mostly local and proportionate, but the skill also directs local command execution and durable logging through env-controlled paths and an unverifiable external redactor/wrapper. That is not enough to call it malicious, but it raises medium trust and privacy risk around the surrounding support tooling rather than the PostHog audit itself.
Confidence: 83%Severity: 53%
Audit Metadata