--- title: KASS: AI Agents Are Moving From Smart-Contract Findings... description: A newly published research paper introduces KASS (Knowledge-Augmented Attack Synthesis and Simulation), an AI agent pipeline that turns smart-contract... canonical: https://daily.dev/posts/kass-ai-agents-are-moving-from-smart-contract-findings-to-executable-exploits-6dinihaqv twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: KASS: AI Agents Are Moving From Smart-Contract Findings to Executable Exploits | daily.dev og:description: A newly published research paper introduces KASS (Knowledge-Augmented Attack Synthesis and Simulation), an AI agent pipeline that turns smart-contract... og:url: https://daily.dev/posts/kass-ai-agents-are-moving-from-smart-contract-findings-to-executable-exploits-6dinihaqv og:image: https://api.daily.dev/og/posts/6dIniHaQV.png og:image:alt: KASS: AI Agents Are Moving From Smart-Contract Findings to Executable Exploits og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # KASS: AI Agents Are Moving From Smart-Contract Findings to Executable Exploits **[Coins Bench](https://daily.dev/sources/coinsbench)** · 6 min read · 1 upvotes · 0 comments ## Summary A newly published research paper introduces KASS (Knowledge-Augmented Attack Synthesis and Simulation), an AI agent pipeline that turns smart-contract vulnerability findings into executable proof-of-concept exploits using Solidity and Foundry, rather than stopping at static-analysis warnings. Using GPT-5.1 on 104 SmartBugs-Curated contracts, KASS produced working PoCs for 94.23% of cases, but ablation tests show most of that success comes from an internal code-repair retry loop (success drops to 22.4% without it) and outer replanning. The system also reproduced 9 of 11 real CVEs with DeepSeek-V3.2, demonstrated a BEC Token integer-overflow exploit with quantified damage, failed on a time-dependent reentrancy case because it didn't manipulate block timestamps, and in one case correctly judged a Slither-flagged reentrancy warning as infeasible due to gas stipend constraints. The piece closes with defensive recommendations: requiring executable PoCs for high-severity findings, using Foundry's invariant and fork testing, testing time/block-dependent state explicitly, and layering access controls like OpenZeppelin's TimelockController and Pausable as containment. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What success rate did the KASS AI agent achieve generating executable smart contract exploits with GPT-5.1? KASS produced successful executable proof-of-concept exploits for 98 of 104 contracts from the SmartBugs-Curated dataset, a 94.23% success rate, using a GPT-5.1 configuration. The benchmark covered reentrancy, arithmetic, denial of service, and unchecked-call vulnerability categories. Removing the inner code-repair retry loop dropped success to only 22.4%, showing the retry mechanism drives most of the result. _Teams evaluating agentic exploit-generation tools can track benchmarks like this one on daily.dev._ ### Why did the KASS exploit-generation agent fail on a smart contract with a timelock reentrancy pattern? KASS failed on a WALLET contract because it generated the deposit and withdrawal calls inside the same transaction, so the required condition that the current timestamp exceed unlockTime was never satisfied. The fix would have required advancing blockchain time between the deposit and withdrawal steps, using something like vm.warp, which the agent did not produce. _Auditors testing time-dependent contract logic can follow reproducible failure cases like this via daily.dev._ ### How does KASS decide whether a Slither-reported reentrancy warning is actually exploitable? KASS attempted to construct a real attack from a Slither reentrancy warning where state was updated after a transfer() call, resembling a checks-effects-interactions violation. Its planner accounted for the 2,300-gas stipend that transfer() provides and determined that profitable reentrancy through that path was infeasible, effectively overriding the static-analysis finding with an executed simulation. _Developers deciding how much to trust static analyzers like Slither can compare exploit-validation approaches on daily.dev._ ## Similar posts on daily.dev - [From Theory to Agent Team: Building a CrewAI-Powered Solidity Security Auditor](https://daily.dev/posts/from-theory-to-agent-team-building-a-crewai-powered-solidity-security-auditor-aht9fongd) · Coins Bench · 1 upvotes · 0 comments - [Purpose-built AI Security Agent Detected 92% of DeFi Contracts Vulnerabilities](https://daily.dev/posts/purpose-built-ai-security-agent-detected-92-of-defi-contracts-vulnerabilities-1sk4jyywp) · Security Boulevard · 0 upvotes · 0 comments - [Repeated VM Escapes By GPT-5.6-Cyber Based Agents Prove VMs and OS' Require Better Maintenance](https://daily.dev/posts/repeated-vm-escapes-by-gpt-5-6-cyber-based-agents-prove-vms-and-os-require-better-maintenance-xbq0n2l3b) · InfoQ · 0 upvotes · 0 comments --- Tags: [#security](https://daily.dev/tags/security), [#ai-agents](https://daily.dev/tags/ai-agents), [#smart-contracts](https://daily.dev/tags/smart-contracts), [#solidity](https://daily.dev/tags/solidity) [View this post on daily.dev](https://daily.dev/posts/kass-ai-agents-are-moving-from-smart-contract-findings-to-executable-exploits-6dinihaqv) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"KASS: AI Agents Are Moving From Smart-Contract Findings to Executable Exploits","url":"https://daily.dev/posts/kass-ai-agents-are-moving-from-smart-contract-findings-to-executable-exploits-6dinihaqv","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/kass-ai-agents-are-moving-from-smart-contract-findings-to-executable-exploits-6dinihaqv"},"datePublished":"2026-08-12T13:19:46.416Z","dateModified":"2026-09-13T20:58:02.260Z","description":"A newly published research paper introduces KASS (Knowledge-Augmented Attack Synthesis and Simulation), an AI agent pipeline that turns smart-contract...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f7b951d8ea2f567ef197b9fbb4a6ad29?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f7b951d8ea2f567ef197b9fbb4a6ad29?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Coins Bench","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Coins Bench","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/ba340d3c8caa4c2586911332928e24d9","url":"https://daily.dev/sources/coinsbench"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/kass-ai-agents-are-moving-from-smart-contract-findings-to-executable-exploits-6dinihaqv","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"security,ai-agents,smart-contracts,solidity","timeRequired":"PT6M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Coins Bench","item":"https://daily.dev/sources/coinsbench"},{"@type":"ListItem","position":3,"name":"KASS: AI Agents Are Moving From Smart-Contract Findings to Executable Exploits"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/kass-ai-agents-are-moving-from-smart-contract-findings-to-executable-exploits-6dinihaqv#faq","mainEntity":[{"@type":"Question","name":"What success rate did the KASS AI agent achieve generating executable smart contract exploits with GPT-5.1?","acceptedAnswer":{"@type":"Answer","text":"KASS produced successful executable proof-of-concept exploits for 98 of 104 contracts from the SmartBugs-Curated dataset, a 94.23% success rate, using a GPT-5.1 configuration. The benchmark covered reentrancy, arithmetic, denial of service, and unchecked-call vulnerability categories. Removing the inner code-repair retry loop dropped success to only 22.4%, showing the retry mechanism drives most of the result. Teams evaluating agentic exploit-generation tools can track benchmarks like this one on daily.dev."}},{"@type":"Question","name":"Why did the KASS exploit-generation agent fail on a smart contract with a timelock reentrancy pattern?","acceptedAnswer":{"@type":"Answer","text":"KASS failed on a WALLET contract because it generated the deposit and withdrawal calls inside the same transaction, so the required condition that the current timestamp exceed unlockTime was never satisfied. The fix would have required advancing blockchain time between the deposit and withdrawal steps, using something like vm.warp, which the agent did not produce. Auditors testing time-dependent contract logic can follow reproducible failure cases like this via daily.dev."}},{"@type":"Question","name":"How does KASS decide whether a Slither-reported reentrancy warning is actually exploitable?","acceptedAnswer":{"@type":"Answer","text":"KASS attempted to construct a real attack from a Slither reentrancy warning where state was updated after a transfer() call, resembling a checks-effects-interactions violation. Its planner accounted for the 2,300-gas stipend that transfer() provides and determined that profitable reentrancy through that path was infeasible, effectively overriding the static-analysis finding with an executed simulation. Developers deciding how much to trust static analyzers like Slither can compare exploit-validation approaches on daily.dev."}}]} ```