antidote-threat-handler

Ideological drift detection identifies shifts in AI responses toward unsafe patterns like premise abandonment or warm acceptance of harmful requests. This Skill monitors conversational interactions continuously, flags drift indicators in real time, and classifies threat severity to enable corrective action before misalignment escalates.

Threat tiers range from 1–4, scaling from minor drift to critical misalignment. Each tier triggers pre-defined corrective protocols—specific antidote responses designed to restore alignment and safety. The Skill applies tier-matched corrective actions automatically based on detected drift indicators.

Yes. This Skill applies to any conversational AI regardless of architecture or framework. It implements continuous drift monitoring, threat classification, and corrective response protocols across different AI interaction types and platforms.

The protocol detects warm acceptance of unsafe premises, premise abandonment in conversation flow, and sycophantic validation patterns. These indicators signal ideological drift and trigger threat classification to assess misalignment severity.

No external dependencies are required. The Skill operates standalone with no prerequisite tools or environment configuration. You can run the Antidote Protocol scan directly on any conversational interaction.

Critical threats (tier 4) trigger the strongest corrective protocol to immediately restore alignment and safety. Lower tiers apply proportional antidote responses. All corrective actions are tier-based and designed to counter unsafe tendencies before they compound.