🧰 GPT-5

The items that name it 3

• ToolLeak: Tool-Invocation Attacks on Coding Agents
  paperas affected products2026-09-21

  used as backend in emulation experiments

• XBOW: GPT-5.5 and Opus 4.7 in Offensive Security Workflows
  writeupas tools2026-05-06

  LLM used as a comparison baseline; missed 40% of vulnerabilities on XBOW's benchmark

• Agent Capability Is a System Design Problem: Lessons From a 90% Improvement on CyberGym | depthfirst
  articleas affected products2025-11-24It is written here as GPT-5 (baseline about 30% on depthfirst's dev set).

  baseline about 30% on depthfirst's dev set

In every workspace you can read 12