vlm-verifier

What problem does it solve?

It solves the problem of inconsistent or potentially falsified enterprise credit application materials by cross-checking evidence across images, documents, and transaction data, then producing an explainable anti-fraud verification report with clear confidence and audit readiness.

Core Features & Use Cases

• Cross-modal evidence verification: Parses VLM-relevant image evidence and correlates it with LLM-based analysis of textual and structured inputs to validate facts across sources.
• Detection-point workflow with iterative reasoning: Builds independent detection points, performs Think→Check→Research loops, and records when results are confirmed vs. pending.
• Compliance-first, audit-traceable reporting: Enforces strict constraints (no speculation, no approval/risk-judgment language), applies data de-identification rules, and generates a JSON audit trail.

Quick Start

Use the vlm-verifier skill to verify whether the uploaded enterprise materials contain cross-modal inconsistencies for anti-fraud review, using your input files and the provided references to generate the structured verification report.