Works with the AI tools you already use

    CClaude CodeCCursorCCodex CLIGGitHub CopilotGGemini CLI+17 more

    Payment Fraud Triage — BEC & Vendor Fraud Verdict (CLEAR/VERIFY/HOLD)

    by Julian GM

    2

    Triage a vendor email, invoice, or bank-detail change request for BEC/fraud signals before you pay — mechanical CLEAR / VERIFY / HOLD verdict, honest about what it can't check.

    Secure checkout via Stripe

    0 installsSecurity scanned

    See it in action

    You say

    Verify this email from acme-corp.co. They say their IBAN changed for the latest invoice. Here is the message and our master data for acme-corp.com.

    Your agent does

    Verdict: HOLD Facts: Sending domain is acme-corp.co (Known: acme-corp.com). IBAN changed to DE91... Signals: F2 (Lookalike domain), F1 (Unverified account change). What moves this to CLEAR: Independent callback to the phone number in master data (not the email) to confirm the IBAN.

    What you get

    Audit bank account change requests for vendor impersonation signals.Identify homoglyph and lookalike domains used in BEC attacks.Detect thread hijacking by analyzing inconsistencies in quoted history.Enforce independent verification standards for all payment detail updates.

    About this skill

    What it does A method the AI executes over a payment artifact you supply — a vendor email, an invoice, a bank-detail change request, a wire/ACH instruction — and checks it against the fraud patterns that actually cost money: business email compromise (BEC), vendor impersonation, lookalike/homoglyph domains, thread hijacking, CEO fraud, unauthorized IBAN/account changes. Output is a mechanical verdict: CLEAR, VERIFY, or HOLD — plus, on every HOLD or VERIFY, the exact action that would move it to CLEAR. The core question this skill is built around: when a vendor says "our bank account changed," who actually confirmed that, and where did the confirmation number come from? Most fraud losses in this category don't come from a badly-worded email — they come from a friendly, confident narration next to one hard fact nobody re-checked. Honest by design Every decisive claim is labeled verified, reported, or unverified. A control that wasn't executed is unverified — never quietly upgraded to reported because the story sounded convincing. Unverified sender does not equal verified sender. "We called and they confirmed" is not evidence unless the number called came from your own vendor records, not from the email being checked — this skill is built specifically to catch that distinction, because "not-checked" and "no hits" are two different claims and this method never blurs them into one. It also refuses to become something it isn't: no drafting phishing lures or "test" social-engineering content, and no profiling the employee who happened to send the email — the subject is always the transaction and the counterparty entity, never a person. Audited & signed Ships with a GPG-signed integrity manifest (MANIFEST.sha256 + MANIFEST.sha256.asc) covering the full assessment logic — the signal matrix, the verdict formula, the anti-false-CLEAR and anti-false-HOLD rules. Verify it yourself in three commands (references/VERIFY.md). Cross-check the fingerprint against this listing — an independent channel, exactly as the verification instructions themselves require:

    3E5C 01D0 6619 A997 5C83 DDA7 7273 DD27 D9C1 E2B1

    What's inside

    How to install

    Drop the file into your AI Agent. Works with Claude, Cursor, ChatGPT, and 20+ more.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    30-day refund

    Not a fit? Get your money back

    Trust & safety

    Security scanned

    Verified clean 27 days ago

    Listed27 days ago

    Creator

    Julian GM

    Julian GM

    4 skills on Agensi

    Frequently Asked Questions

    Popular in Business & Operations