CLAIM-2026-09-06-02

The disclosure covers eight findings across seven AI coding agents: Claude Code (two findings), OpenAI Codex, Cursor, goose, Qwen Code, Grok Build and Hermes Agent.

Why it matters

Seven agents means this is a class, not one vendor's bug. If you standardised on one agent to reduce risk, you did not reduce this one.

Sources

  • manifoldsec2026gitspawn
  • thehackerne2026malicious