Gen adds Agent Trust Hub safety ratings to Vercel’s skills.sh

Gen adds independent safety ratings to Vercel’s AI skills directory

Gen is embedding its Agent Trust Hub (ATH) into Vercel’s skills.sh open directory to bring independent, AI-driven safety verification and transparent risk ratings to the expanding market for modular AI agent skills. Under the collaboration, each skill published on skills.sh is analysed by ATH and assigned one of four classifications — Safe, Low Risk, High Risk or Critical Risk — so developers and end users see clearer security posture information before installing or executing agent capabilities.

The move addresses growing concerns as AI agents evolve from simple assistants into autonomous digital actors that browse the web, call APIs, access sensitive data and trigger multi-step workflows. Until now, safety assessments often rely on limited metadata or community reputation, leaving organizations exposed to poorly designed or malicious skills that can introduce vulnerabilities, leak data or enable fraud. Gen says embedding independent verification directly into the discovery layer aims to scale innovation while reducing those risks.

ATH’s assessments and reasoning are surfaced alongside each skills.sh listing, including machine‑readable metadata, audit trails and suggested mitigations to lower deployment risk. Gen and Vercel say the system is designed to reduce incidents caused by exploitable skills, accelerate safe adoption of composable capabilities and provide an auditable record for organizations and regulators evaluating third‑party agent components. The companies expect a pilot rollout later this year. Howie Xu, Gen’s chief AI and innovation officer, says trust must be built into the ecosystem as agents gain agency, while Vercel’s chief of software, Andrew Qu, says the integration helps developers move quickly with greater transparency for users.

Vercel operates skills.sh as a discovery layer where developers publish and install modular skills that extend autonomous agents’ capabilities, and its platform serves more than 6 million developers worldwide. The integration therefore exposes ATH’s independent assessments to a broad developer base, aiming to influence compliance and security practices at the point of adoption.

The collaboration signals a broader industry shift toward standardising safety and auditability in marketplaces for composable AI components. For Gen, which positions itself at the intersection of cybersecurity and AI safety, the partnership with a major developer platform represents a strategic effort to set verification norms that could shape regulation, procurement and enterprise trust in agent-driven software.