Show HN: Xtra – a Python framework for reasoning about AI system threats

Xtra, a Python framework from Tangible Research, detects conversational social engineering using a finite state machine that models conversation state transitions without LLMs or embeddings. It identifies attacks by analyzing patterns like flattery density, asymmetry spikes, and escalation velocity, flagging extraction attempts based on path through state space.

Conversational social engineering detection using a Finite State Machine. No LLMs. No embeddings. No semantic search. Pure algorithmic detection. xtra models conversations as state transitions: NEUTRAL → RAPPORT BUILD → TRUST ESTABLISHED → EXTRACTION ATTEMPT → FLAGGED Attacks are detected by their path through state space, not their content. - Flattery density and momentum across turns - Asymmetry spike give/ask ratio collapse - Escalation velocity - Reciprocity pressure - Decoy turn detection - Scope mismatch python from xtra import Xtra, TurnSignals result = Xtra .analyze turns if result.is attack: print result.flag reason, result.confidence Making AI Tangible — github.com/TangibleResearch