Google Fixes CVSS 10 Gemini CLI CI RCE and Cursor Flaws Enable Code Execution
Summary: Google patched a CVSS 10.0 flaw, with no CVE identifier, in the @google/gemini-cli npm package and the google-github-actions/run-gemini-cli workflow that let an unprivileged external attacker load malicious content as Gemini configuration and execute commands on the host before the agent's sandbox initialized. The issue was limited to headless CI use, where Gemini CLI automatically trusted workspace folders and loaded configuration from a local .gemini/ directory; version 0.39.1 now requires explicit folder trust and evaluates tool allowlisting under --yolo mode.
Why it matters: Pillar Security's Dan Lisichkin described the allowlisting bypass as a path from a single GitHub issue, through indirect prompt injection, to theft of the build environment's GITHUB_TOKEN and a push to the gemini-cli main branch - and noted that an allowlist permitting run_shell_command(cat) still leaves /proc and .git/config reachable.
Read the original at The Hacker News ↗
- Topics
- Supply Chain & AppSec
- Related
- Kitesurf is now open source Cyber Risks · Nov 13, 2026Trained to betray: How post-training supply chain attacks hijack coding agents Cyber Risks · Nov 12, 2026ZDI-26-913: (0Day) Meta fairseq load_audio_path Code Injection Remote Code Execution Vulnerability Cyber Risks · Nov 11, 2026