Expired50% confidenceFactExact time
Researchers analyzed more than 42,000 AI Agent Skills, of which 26.1% had security vulnerabilities and 5.2% showed suspected malicious signs.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
SkillSpector: NVIDIA's Open-Source Security Scanner for AI Agent Skills
bilibiliAIlazy俊6/19/2026
Related Claims
Unverified2024 年多家安全研究机构发现，自主 AI Agent 在执行文件操作时可能因上下文理解偏差导致误删关键文件，或在依赖安装过程中引入存在漏洞的第三方包75% similarVerifiedA 2022 Stanford University study found that developers using AI coding assistants had a significantly higher probability of writing code containing security vulnerabilities compared to a control group that did not use AI.74% similarUnverifiedA creator with 10 years of AI research experience, a GitHub tutorial repository with a cumulative 83K stars, and a newsletter with over 40,000 subscribers found through research that the biggest pain point for developers using AI programming is the issue of trust.73% similarUnverifiedMcKinsey research shows that enterprises with the fastest adoption of AI tools face three major internal challenges: accelerating skills stratification, loss of quality control, and data security risks.72% similarUnverified2023年安全研究机构Vulcan Cyber的研究报告显示，主流AI编程助手中幻觉包名的出现率在特定场景下高达21.3%72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/41753API
curl https://kongchang.com/api/v1/knowledge/claims/41753MCP
get_claim(id=41753)