Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    invoice expense sanity gate Skill Pro

    1

    An adversarial review agent that flags arithmetic errors, duplicates, and policy violations in financial documents.

    $9

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    invoice expense sanity gate Skill Pro

    invoice expense sanity gate Skill Pro

    Example session with this skill installed

    Review this invoice: Vendor: Acme Corp, Inv #1024, Date: 2023-10-01. Items: 10x Widget A @ $15.00, 5x Widget B @ $20.00. Tax: $20.00. Total: $280.00. PO #882 states Widget A price is $12.00.

    • Read your context and instructions
    • Compiled the invoice expense sanity

    I have completed the review of Invoice #1024 from Acme Corp. Several discrepancies were found regarding pricing and arithmetic.

    Invoice Review Report

    Decision Snapshot

    • Status: HOLD
    • Confidence: High
    • Document Type: Invoice
    • Reference: #1024
    • Vendor: Acme Corp
    • Amount: $280.00 USD

    Extracted Data

    FieldValueConfidence
    VendorAcme CorpHigh
    Inv Number1024High
    Date2023-10-01High
    PO Number882 (from input)High

    Arithmetic Reconciliation

    • Stated Total: $280.00
    • Recalculated Subtotal: (10 * 15.00) + (5 * 20.00) = $250.00
    • Recalculated Grand Total: $250.00 + $20.00 (Tax) = $270.00
    • Variance: $10.00 (Overcharged on invoice)

    Red Flags & Evidence

    • Severity: High | Arithmetic Mismatch | Stated total ($280) does not match sum of parts ($270).
    • Severity: Medium | Price Variance | Widget A billed at $15.00; PO #882 specifies $12.00.
    • Severity: Low | Missing Info | No payment terms or due date extracted from provided text.

    Policy & Approval Check

    • Rule Tested: PO Matching
    • Result: Exception found. Price per unit for Widget A exceeds authorized PO amount by $3.00.

    Recommended Actions

    • Contact Acme Corp to resolve the $10.00 arithmetic error.
    • Verify if a price increase for Widget A was approved outside of PO #882.
    • Request a corrected invoice reflecting the agreed contract price.

    Next steps

    1. Cross-check Widget A pricing against the master service agreement.
    2. Flag this invoice in the accounting system to prevent automated payment.
    3. Request the full PDF to check for missing payment terms.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Accounts payable and expense management often rely on manual checks that miss small errors or duplicate submissions. This skill acts as an adversarial review layer for invoices, receipts, and vendor bills. It identifies arithmetic mismatches, missing evidence, and potential duplicate payments before they hit your ledger.

    What it does

    • Data extraction captures vendor details, line items, taxes, and reference numbers from text or images.
    • Arithmetic reconciliation recalculates totals from unit prices and quantities to flag rounding or sum errors.
    • Duplicate detection compares current submissions against historical records to identify near-matches or split transactions.
    • Evidence verification checks for missing support like purchase orders, delivery records, or required receipts.
    • Policy auditing tests expenses against provided spending limits, category rules, and approval thresholds.

    How it works

    1. Input documents by providing raw text, images, or files of the invoice or expense report.
    2. Review logic triggers the extraction and normalization of all financial data points.
    3. Validation checks run the arithmetic tests and cross-reference against any provided policies or POs.
    4. Risk assessment generates a report flagging items as CLEAR, REVIEW, or HOLD with prioritized actions.

    Frameworks & tools

    Compatible with any agentic workflow capable of processing text or document images. It uses structured extraction and comparison logic suited for procurement, bookkeeping, and ERP-adjacent tasks.

    Why this beats prompting it yourself

    Standard prompts often assume documents are correct. This skill uses a skeptical, evidence-first approach designed to find inconsistencies that generic LLM prompts overlook, specifically focusing on recalculated variances and cross-document matching.

    Use cases

    • Vendor bill audit to catch overcharges or tax calculation errors before payment.
    • Expense report review to flag policy violations or missing receipts in employee claims.
    • Duplicate prevention to identify invoices submitted multiple times under different filenames.
    • Procurement matching to verify that invoice line items align with original purchase orders.

    Known limitations

    Does not provide legal or tax opinions. Cannot verify the authenticity of a document from appearance alone. Requires manual confirmation of OCR values marked with low confidence.

    Cannot determine document authenticity or intent. Requires manual confirmation for low-confidence OCR. Does not provide tax or legal certification. Not a substitute for final human approval.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 18 days ago

    Permissions

    No special permissions declared or detected

    Listed18 days ago

    What's inside

    Frequently Asked Questions

    This skill acts as an adversarial auditor that uses recalculation and historical cross-referencing to find errors, duplicates, and policy breaches that generic LLMs often miss due to their tendency to assume document accuracy.

    The skill accepts raw text, document images (OCR), and structured files like PDFs, making it compatible with any agentic workflow capable of processing financial document data.

    You will receive the structured review logic, extraction schemas, and the risk assessment framework, along with documentation on how to integrate it into your existing procurement or bookkeeping workflow.

    While it flags discrepancies and recalculates sums, it cannot verify if a document is a legally forged physical copy or provide professional tax advice; a human should always perform a final review of flagged items.

    No, the skill includes a duplicate detection layer that compares new submissions against your historical records to identify near-matches, split transactions, or re-submissions under different filenames.

    Buyers receive lifetime access to the version purchased, including any logic refinements and schema updates released by the creator to improve extraction accuracy.

    New in Finance & Accounting