Accident Investigation Prover Connector for AI agents.

It forces your AI to follow ICAO Annex 13 standards, ensuring you don't miss systemic issues like maintenance logs or organizational pressures.

Yes, it uses Reason's Model and the HFACS taxonomy to trace active failures back to latent conditions like budget constraints or training gaps.

It requires the AI to correlate FDR parameters and CVR transcripts with ATC tracks and maintenance records before it'll allow a conclusion.

It rejects vague goals like 'improve training' and forces the agent to write specific, measurable, and evidence-linked actions for named authorities.

Yes, it's built specifically to enforce the rigor required by NTSB and ICAO for multi-causal accident analysis.

It can help you analyze MEL deferral rates and component life limits to see if they contributed to a specific incident.

The engine maintains a semantic trap list of blame-language signals: 'pilot error,' 'crew error,' 'human error,' 'judgment error,' 'failed to,' 'negligence.' If the LLM uses any of these in the HFACS taxonomy field, the classification is rejected. Instead, the LLM must classify each factor at the correct HFACS level: Level 1 (what the pilot did. skill error, decision error, perceptual error, or violation), Level 2 (what conditions enabled it. fatigue, CRM failure, environment), Level 3 (what supervision allowed it. scheduling, training gaps), Level 4 (what organizational decisions created it. budget cuts, staffing, regulatory gaps). 'Pilot error' is Level 1 only. the investigation must reach Level 4.

Five mandatory sources: (1) Flight Data Recorder. minimum 88 parameters per ICAO Annex 6, with timestamps. (2) Cockpit Voice Recorder. last 2 hours of audio, transcribed with timestamps, correlated with FDR parameter changes. (3) ATC recordings. radar track, clearances, handoffs, weather advisories. (4) Maintenance logs. last A/B/C/D checks, MEL items, deferred defects, AD compliance, component life (TSN/TSO/CSN/CSO). (5) Wreckage analysis. impact signatures, fire patterns, fracture analysis (fatigue vs overload), metallurgical examination. The engine rejects speculative language like 'it appears that' or 'evidence suggests' when hard data from these sources exists.

Because 'improve training' with no addressee has zero accountability. The NTSB model requires each recommendation to name the authority responsible for implementation. the FAA, the operator, the manufacturer, or an international body. Each recommendation must include: what specific action to take, measurable success criteria, a response deadline (typically 90 days), a verification mechanism (audit, inspection, data review), and a direct link to a specific investigation finding. This is how aviation achieved its extraordinary safety record. not through vague wishes, but through tracked, accountable, evidence-linked changes.