Checks whether a merchant storefront looks like the same legitimate entity it appeared to be recently, BEFORE an agent commits payment to it.

This is the mirror image of merchant-side agent-verification protocols (e.g. Visa's Trusted Agent Protocol): those let a merchant confirm an incoming agent is legitimate. This tool lets the AGENT confirm the MERCHANT is legitimate before paying.

It does NOT guarantee a storefront is safe - it is a fast, explainable pre-purchase signal built from free public data sources (domain registration history via RDAP, SSL certificate issuance history via Certificate Transparency logs, HTTPS validity, and a known-scam-domain check). Treat a "high" risk result as a strong reason to pause and get human confirmation before paying, and a "low" risk result as one input among others, not a guarantee.

Args:

• domain (string): the bare domain of the storefront to check, e.g. "example-shop.com"

Returns: { "domain": string, "trustScore": number, // 0-100, higher = more trustworthy "riskLevel": "low" | "medium" | "high" | "unknown", "confidence": "low" | "medium" | "high", // how many signals could actually be resolved "reasons": string[], // human-readable explanation for the score "recommendation": "proceed" | "pause_for_confirmation" | "do_not_proceed", "recommendationReason": string, // one-line explanation of why this recommendation was made "signals": { ...raw signal data for audit/debugging... }, "checkedAt": string // ISO timestamp - this is a point-in-time result, do not cache it }

Recommendation values and suggested agent behavior:

• "proceed": trust score is high with high confidence — complete the transaction silently, no user confirmation needed

• "pause_for_confirmation": score is borderline or confidence is insufficient — pause and show recommendationReason to the user before completing payment

• "do_not_proceed": domain matched a scam list or trust score is critically low — block the transaction and actively notify the user (do not fail silently)

Examples:

• Use when: an agent is about to submit payment to a storefront it found via search or a marketplace listing

• Use when: comparing several candidate storefronts for the same product before choosing one to buy from

• Don't use when: verifying a well-known, previously-transacted-with merchant on every single repeat purchase - reserve for new or unfamiliar storefronts, since results are not meant to be cached long-term but also aren't needed for every routine repeat transaction

Error Handling:

• Never throws - if a data source is unreachable, the affected signal is returned as null and "confidence" drops accordingly rather than failing the call