Brief

Researchers have developed AgentScore, a novel method for automatically constructing clinical scoring systems using LLM agents. This approach addresses the challenge of creating interpretable and deployable clinical guidelines by leveraging LLMs to propose rules and a verification loop to ensure statistical validity. AgentScore demonstrated superior performance compared to existing methods across eight clinical prediction tasks and outperformed established scores on two external validation tasks. AI

A security scanner named AgentScore, designed to detect command injection vulnerabilities in npm packages, underwent four rounds of iterative refinement over a 96-hour period in mid-May 2026. Initially, the scanner flagged 31 packages, leading to hypotheses of widespread developer error or scanner over-sensitivity. Through manual audits and the development of new context-aware mitigators, the scanner was improved to better distinguish between genuine threats and benign code patterns, such as internal helper paths or SQL queries. AI