Novee Security showed at Black Hat that a GitHub issue opened by an account with no repository access could reach the CI runners behind major AI coding agents in their default configurations, tested against Claude Code, Gemini CLI, and Codex. The strongest, a Gemini CLI container-launcher command injection scored 10.0, runs code on the CI host before the sandbox starts. In Claude Code, a validator that stripped quoted text let a payload in a Git flag reach the runner, and a separate flaw leaked an API key through a download counter. Untrusted issue content reaching an agent that holds secrets and tools in the same runtime is the shared weakness.
Researchers at Pillar Security demonstrated sandbox escapes across four widely used AI coding agents: Cursor, OpenAI Codex CLI, Google Gemini CLI, and Antigravity. In nearly every case the agent never broke the sandbox directly; it only had to write a file that a trusted component outside the sandbox would later run, load, or scan. Failure modes included hook abuse, editing a virtual environment interpreter the editor then ran itself, planting Git metadata outside a .git folder to fire execution through fsmonitor, and a command allowlist that trusted a tool by name while the real invocation was not read only. Prompt injection in workspace content was the trigger.
Researchers at Mozilla's 0DIN found that an AI coding agent told to clone and set up a seemingly harmless GitHub repository can be tricked into running malware that stays invisible to security scanners, the agent itself, and human reviewers. The trick is that nothing malicious sits in the repository's files. Instead, a routine-looking setup command runs a script that fetches a value hidden in a DNS TXT record and executes it as a shell command, pulling down and running an attacker's payload like a reverse shell. Because the payload lives outside the repo and arrives over DNS at setup time, code review and static scanning see nothing wrong.
TrustStrike Labs