--- title: “Urgency is an inherently subjective call”: Anthropic... description: Anthropic explained to The New Stack how it tests Claude Duet, an AI agent that monitors connected apps and flags priorities without being asked. Internal... canonical: https://daily.dev/posts/urgency-is-an-inherently-subjective-call-anthropic-tells-tns-how-it-grades-claude-duet-xyeheykc0 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: “Urgency is an inherently subjective call”: Anthropic tells TNS how it grades Claude Duet | daily.dev og:description: Anthropic explained to The New Stack how it tests Claude Duet, an AI agent that monitors connected apps and flags priorities without being asked. Internal... og:url: https://daily.dev/posts/urgency-is-an-inherently-subjective-call-anthropic-tells-tns-how-it-grades-claude-duet-xyeheykc0 og:image: https://api.daily.dev/og/posts/xyeHEykc0.png og:image:alt: “Urgency is an inherently subjective call”: Anthropic tells TNS how it grades Claude Duet 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. # “Urgency is an inherently subjective call”: Anthropic tells TNS how it grades Claude Duet **[The New Stack](https://daily.dev/sources/newstack)** · 5 min read · 0 upvotes · 0 comments ## Summary Anthropic explained to The New Stack how it tests Claude Duet, an AI agent that monitors connected apps and flags priorities without being asked. Internal graders assess task prioritization accuracy, but Anthropic admits 'urgency is an inherently subjective call' and the eval is imperfect. A separate automated check, built on Claude Code's auto mode classifier, assesses whether each action matches user intent and runs independently so hidden webpage instructions can't directly steer it. Duet, powered by Claude Opus 5.5, checks connected apps every few hours rather than continuously, partly to conserve a subscriber's five-hour session and weekly usage limits. The research preview is available by waitlist to U.S. Pro and Max subscribers. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How does Claude Duet prevent prompt injection from webpages from hijacking its actions? Claude Duet runs a separate, automated classifier alongside each action that checks whether the action matches the user's actual intent. This classifier is built on the auto mode classifier from Claude Code and runs independently of the model performing the task, so instructions hidden in a webpage cannot directly steer its decision. _Developers building agent safeguards against prompt injection can track real-world agent security approaches on daily.dev._ ### Does Claude Duet's background monitoring count against my Claude usage limits? Yes, Claude Duet uses the same allowance as regular Claude conversations, so background work counts against a subscriber's five-hour session and weekly plan limits. Larger tasks like building a financial model or deck require more model turns and background work, and even routine checks across connected apps count toward those limits. _Anyone weighing AI agent subscriptions against usage caps can follow agent pricing details on daily.dev._ ### How often does Claude Duet check connected apps for urgent items? Claude Duet checks connected apps every few hours rather than continuously, partly to avoid exhausting a subscriber's usage allowance. This means something urgent could arrive between checks and go unnoticed until the next scheduled check, a tradeoff Anthropic made between resource usage and alert latency. _Teams relying on background AI agents for alerts can track these tradeoffs via daily.dev._ ## Similar posts on daily.dev - [Claude wants to rule your personal life now](https://daily.dev/posts/claude-wants-to-rule-your-personal-life-now-8nxoeoizx) · XDA Developers · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#claude](https://daily.dev/tags/claude), [#anthropic](https://daily.dev/tags/anthropic), [#prompt-injection](https://daily.dev/tags/prompt-injection) [View this post on daily.dev](https://daily.dev/posts/urgency-is-an-inherently-subjective-call-anthropic-tells-tns-how-it-grades-claude-duet-xyeheykc0) ```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":"“Urgency is an inherently subjective call”: Anthropic tells TNS how it grades Claude Duet","url":"https://daily.dev/posts/urgency-is-an-inherently-subjective-call-anthropic-tells-tns-how-it-grades-claude-duet-xyeheykc0","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/urgency-is-an-inherently-subjective-call-anthropic-tells-tns-how-it-grades-claude-duet-xyeheykc0"},"datePublished":"2026-10-29T11:01:52.995Z","dateModified":"2026-10-29T22:07:55.766Z","description":"Anthropic explained to The New Stack how it tests Claude Duet, an AI agent that monitors connected apps and flags priorities without being asked. Internal...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/410689613f2ba9d8dbf77fdff937be0d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/410689613f2ba9d8dbf77fdff937be0d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The New Stack","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":"The New Stack","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/newstack","url":"https://daily.dev/sources/newstack"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/urgency-is-an-inherently-subjective-call-anthropic-tells-tns-how-it-grades-claude-duet-xyeheykc0","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,claude,anthropic,prompt-injection","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The New Stack","item":"https://daily.dev/sources/newstack"},{"@type":"ListItem","position":3,"name":"“Urgency is an inherently subjective call”: Anthropic tells TNS how it grades Claude Duet"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/urgency-is-an-inherently-subjective-call-anthropic-tells-tns-how-it-grades-claude-duet-xyeheykc0#faq","mainEntity":[{"@type":"Question","name":"How does Claude Duet prevent prompt injection from webpages from hijacking its actions?","acceptedAnswer":{"@type":"Answer","text":"Claude Duet runs a separate, automated classifier alongside each action that checks whether the action matches the user's actual intent. This classifier is built on the auto mode classifier from Claude Code and runs independently of the model performing the task, so instructions hidden in a webpage cannot directly steer its decision. Developers building agent safeguards against prompt injection can track real-world agent security approaches on daily.dev."}},{"@type":"Question","name":"Does Claude Duet's background monitoring count against my Claude usage limits?","acceptedAnswer":{"@type":"Answer","text":"Yes, Claude Duet uses the same allowance as regular Claude conversations, so background work counts against a subscriber's five-hour session and weekly plan limits. Larger tasks like building a financial model or deck require more model turns and background work, and even routine checks across connected apps count toward those limits. Anyone weighing AI agent subscriptions against usage caps can follow agent pricing details on daily.dev."}},{"@type":"Question","name":"How often does Claude Duet check connected apps for urgent items?","acceptedAnswer":{"@type":"Answer","text":"Claude Duet checks connected apps every few hours rather than continuously, partly to avoid exhausting a subscriber's usage allowance. This means something urgent could arrive between checks and go unnoticed until the next scheduled check, a tradeoff Anthropic made between resource usage and alert latency. Teams relying on background AI agents for alerts can track these tradeoffs via daily.dev."}}]} ```