{"product_id":"detect-hidden-prompt-injections-in-cvs-with-gemini-drive-sheets","title":"Detect Hidden Prompt Injections in CVs with Gemini, Drive \u0026 Sheets","description":"\u003ch3\u003eDetect Hidden Prompt Injections in CVs—Automatically with Gemini, Google Drive \u0026amp; Sheets\u003c\/h3\u003e\n\u003cp\u003eThis n8n workflow screens uploaded CV PDFs for \u003cstrong\u003ehidden-text and prompt-injection attempts\u003c\/strong\u003e by comparing Google Gemini’s text extraction verdict with Gemini Vision’s image-based verdict—then quarantines suspicious files in \u003cstrong\u003eGoogle Drive\u003c\/strong\u003e and logs results to \u003cstrong\u003eGoogle Sheets\u003c\/strong\u003e.\u003c\/p\u003e\n\n\u003ch3\u003eWhat this workflow does\u003c\/h3\u003e\n\u003cul\u003e\n  \u003cli\u003e\n\u003cstrong\u003eReceives applications\u003c\/strong\u003e via an n8n Form with candidate name, selected role, and a \u003cstrong\u003ePDF CV upload\u003c\/strong\u003e.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eValidates file type\u003c\/strong\u003e: if the upload is not a PDF, the workflow rejects it and returns a rejection response.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eRuns Gemini on extracted text\u003c\/strong\u003e: the PDF is processed to extract raw text, which is sent to \u003cstrong\u003eGoogle Gemini\u003c\/strong\u003e to generate a structured verdict (qualified flag, score, years of experience, and a one-sentence reason).\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eRuns Gemini Vision on the rendered CV\u003c\/strong\u003e: the PDF is converted to a JPG (via \u003cstrong\u003eConvertAPI\u003c\/strong\u003e), then the image is sent to \u003cstrong\u003eGoogle Gemini vision\u003c\/strong\u003e for the same structured verdict.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eCompares verdicts to detect tampering\u003c\/strong\u003e: it calculates divergence between text vs. vision outputs and flags the CV as suspicious if they disagree or exceed configured thresholds (including detection of hidden-text \/ prompt-injection attempts).\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eQuarantines and alerts\u003c\/strong\u003e: if flagged, the original CV is uploaded to a \u003cstrong\u003eGoogle Drive quarantine folder\u003c\/strong\u003e and a detailed \u003cstrong\u003eTelegram\u003c\/strong\u003e alert is sent. If clean, it sends a “clean result” message using the vision-based reason.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eLogs everything\u003c\/strong\u003e: appends a row to \u003cstrong\u003eGoogle Sheets\u003c\/strong\u003e (the “Applications” tab) with submission metadata, both scores, and divergence metrics.\u003c\/li\u003e\n\u003c\/ul\u003e\n\n\u003ch3\u003eUse cases\u003c\/h3\u003e\n\u003cul\u003e\n  \u003cli\u003eSaaS HR pipelines that need automated detection of manipulated or hidden instructions inside applicant CVs.\u003c\/li\u003e\n  \u003cli\u003eTeams using n8n to protect Gemini-based resume screening from prompt-injection risks.\u003c\/li\u003e\n  \u003cli\u003eOperations workflows requiring auditable reporting in Google Sheets and instant Telegram notifications.\u003c\/li\u003e\n\u003c\/ul\u003e\n\n\u003ch3\u003eTechnical details\u003c\/h3\u003e\n\u003cul\u003e\n  \u003cli\u003e\n\u003cstrong\u003en8n nodes\u003c\/strong\u003e: if, set, form, merge, telegram, sticky note.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eIntegrations\u003c\/strong\u003e: Google Gemini (text + vision), Google Drive (quarantine), Google Sheets (Applications log), Telegram alerts, and ConvertAPI for PDF-to-JPG rendering.\u003c\/li\u003e\n\u003c\/ul\u003e","brand":"N8N Commerce","offers":[{"title":"Default Title","offer_id":45916244050099,"sku":"N8N-18805","price":34.99,"currency_code":"GBP","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/0749\/6279\/6723\/files\/Tkdci4N0jS--HXz34MX8V_Rn402KdB.png?v=1788013350","url":"https:\/\/buyflowscripts.com\/products\/detect-hidden-prompt-injections-in-cvs-with-gemini-drive-sheets","provider":"N8N Commerce","version":"1.0","type":"link"}