Your best
PROMPT_INJECTION,
meet AgentSense.
Starting 11/17/26, attack our AI agents for a chance to win $10,000
View terms & conditions
Open challenge brief
Your mission
In a controlled environment, make an AgentSense-protected agent leak data or write a system file. Switch AgentSense protection on and off to compare how the agent responds.
The prize
The three eligible participants with the shortest validated times for successfully completing the Challenge will each receive $10,000 USD.
The details
The challenge will run from 11/17/26 at 8:30am EST / 3:30pm ILT to 12/8/26 at 12am EST / 7:00am ILT. No purchase is needed. Scroll down for win conditions or view terms and conditions.
To Win:
01
Complete both levels
A chatbot copilot and a coding agent
02
Use indirect prompt injection
Deliver instructions through content the agent reads. Use one of the predefined user prompts to try and make the agent follow your instructions.
03
Have protection on
Your attack must succeed with AgentSense enabled.



Testing Scope
Keep all testing inside the designated sandbox environment. Do not attempt to access Cato systems, services, or data outside it.
About Cato AgentSense
Cato AgentSense secures AI agents by stopping harmful actions before they execute. It assesses agents in real time to protect the integrity of their decisions, actions, and outputs as they operate autonomously. By tracking agent action back to untrusted instructions, it can stop indirect prompt injection, even when the instruction arrives through a trusted source and the attack pattern is unknown.
Ready to get started?
View terms & conditions
AgentSense Challenge Terms and Conditions
The AgentSense Challenge invites you to test Catoβs AI security technology by attempting to make an AgentSense-protected agent leak fictional test data within a controlled sandbox (the βChallengeβ). These terms govern your use of the sandbox, participation in the Challenge, and eligibility for prizes. The Challenge is organized by Cato Networks Ltd. (βCato,β βwe,β or βusβ). By accepting these terms at registration or by participating in the Challenge, you agree to follow them.
1. Organizer and dates. Cato Networks LTD (βCatoβ) runs this free skill challenge from November 17, 2026 at 8:30am to December 8, 2026 at 12am (United States Eastern Time. No purchase is needed.
2. Who can participate. You must be at least 18 and the legal age of adulthood where you live, and reside and participate in an eligible country. You cannot participate if you are a government employee, or if you reside where your participation is prohibited by law or applicable sanctions, or if disbursement of the prize money would be prohibited by applicable sanctions. Cato employees and their immediate families are not eligible. Enter as an individual using one account and an email address. You are solely responsible for obtaining any permission your employer requires to participate, share materials, and accept a prize.
3. The Challenge objective. Crack AgentSense by causing the protected agent to leak your designated fictional challenge data. A qualifying leak must result from indirect prompt injection: instructions you place in contextual material the agent reads, rather than instructions you give it directly. AgentSense protection must be enabled during the qualifying attempt. Your results must identify the target test data, what counts as a leak, and the permitted output or Cato-controlled destination. Do not send data to any other destination. You may compare results with AgentSense protection off, but those results do not qualify. An apparent success is subject to Catoβs validation at Catoβs sole discretion.
4. Stay inside the sandbox. Cato authorizes you to test only the assigned Challenge environment sandbox within these terms and the Challenge rules provided to you inside the sandbox. You will receive a link to the sandbox via the email address that you submit to Cato at registration, after you have accepted these terms. Activities outside this environment are not authorized and do not count toward the Challenge. Do not test Cato production systems, third-party systems, your employerβs systems, other participantsβ accounts, or the sandboxβs underlying infrastructure. Do not bypass account controls or rate limits, disrupt the service, or interfere with other participants.
5. Protecting sandbox access. Your sandbox access is personal to you. Do not share your account, login credentials, access tokens, or sandbox access with anyone, or allow anyone else to use your account. Keep your credentials secure and use the sandbox only through Catoβs approved access methods. You may attempt to bypass AgentSense only through the prompt injection attacks provided to you inside the sandbox. Do not bypass login or account security of the sandbox, attack the sandboxβs underlying infrastructure, disrupt its operation, or access another participantβs account or environment. If you suspect unauthorized access or that your credentials have been exposed, stop using the affected account.
6. Use non-production test data only. Never upload, enter, connect, use, or attempt to obtain production data, real personal data, or sensitive data through the sandbox or in attack submissions. This includes customer or employer information, confidential business information, health or financial records, and real passwords, keys, or credentials. Use only fictional, non-production test data. Do not connect live accounts or production services. Provide your registration and prize-payment information only through the separate forms Cato designates. If you encounter real or sensitive data, stop immediately and report it to [security contact]. Do not copy, share, or continue accessing it. Follow Catoβs instructions for safe handling and deletion.
7. Submissions and ranking. Submit each claimed success through submission method before the closing time, with the payload, sandbox session identifier, and steps needed to reproduce it. Do not include prohibited data. Cato will review submissions after the challenge closes, using the published objectives and sandbox records to verify the result.
8. Prizes and payment. The three eligible participants with the shortest validated times for successfully completing the Challenge will each receive $10,000USD. Each person may win once. If fewer than three participants qualify, only those qualifying prizes will be awarded. These payments are Challenge prizes, not payment for Feedback, a license, or services. Cato will notify potential winners by email by December 14. Respond within 14 days and provide the identity, eligibility, tax, and payment information reasonably needed to pay you. Cato will send a reminder before treating a prize as unclaimed. If you are ineligible or do not complete these steps by the stated deadline, the next eligible participant will be considered. Prizes will be paid by bank transfer within 30 days after verification and receipt of required documents. You are responsible for taxes you owe on your prize. Cato may withhold taxes and report payments as required by law. No additional Feedback, marketing consent, or promotional services are required to receive a prize.
9. Feedback and challenge results. Your provision of ideas, suggestions, comments, or other feedback about the Challenge or AgentSense to Cato (βFeedbackβ) is voluntary. Cato may also collect and analyze Challenge activity, including submitted prompts, code, attack methods, agent responses, logs, and test results (βChallenge Materialsβ). You give Cato a worldwide, perpetual, irrevocable, nonexclusive, transferable, sublicensable, royalty-free license to use, copy, modify, publish, distribute, and otherwise exploit the Feedback and the Challenge Materials for any purpose, without payment to you. Cato may freely use any knowledge, insights, or techniques learned from the Challenge. Share only materials you have the right to provide. Prize money rewards challenge performance and is not payment for Feedback or Challenge Materials.
10. Sharing results. Until Cato publicly announces the Challenge results, do not publicly share your Challenge results, claimed successes, attack methods, payloads, or other details that would allow someone to reproduce your attempts. This includes social media posts, articles, videos, and public code repositories. You may report findings privately to Cato at any time. This restriction does not prevent disclosures required by law.
11. Your information. Cato uses your registration details, challenge activity, and submissions to operate and secure the Challenge, validate entries, contact you about results, and award prizes. Cato will seek separate permission before using your name, image, or quote in optional publicity, except for disclosures required by law. Catoβs use of personal information is subject to its Privacy Notice, as explained here.
12. Administration. Cato may suspend access or disqualify entries for cheating, false information, or material breaches of these terms. Cato may pause, change, or end the Challenge in Catoβs sole discretion.
13. Participation at your own risk. You participate in the Challenge and use the sandbox at your sole risk. The sandbox is provided βas isβ and βas available.β To the fullest extent permitted by law, Cato disclaims all warranties, whether express, implied, or statutory, including merchantability, fitness for a particular purpose, title, and noninfringement. Cato does not guarantee that the sandbox will be secure, accurate, uninterrupted, or error-free.
14. Limits on liability. To the fullest extent permitted by law, Cato will not be liable for indirect, incidental, special, consequential, exemplary, or punitive damages, or for lost profits, revenue, business opportunities, or data arising from the challenge, sandbox, or these terms. Catoβs total liability to you for all claims combined will not exceed $100USD, regardless of the legal basis of the claim or whether Cato was advised that such losses were possible. This cap does not reduce any prize properly owed to you under these terms.
15. Governing law. These terms are governed by the laws of the State of Israel. The courts of Tel Aviv, Israel shall have exclusive jurisdiction and venue to adjudicate any dispute arising out of these terms.