Claude Code and Gemini CLI Flaws Let a GitHub Issue Reach CI Workflow Secrets
The Hacker News reported that Claude Code and Gemini CLI contain flaws that let a GitHub issue reach CI workflow secrets. The vulnerabilities were discovered by security researchers, though the specific researchers were not named in the excerpt. The attack vector involves a GitHub issue, which suggests that an attacker could craft a malicious issue to trigger the AI tools to expose secrets. The source states that these flaws allow a GitHub issue to reach CI workflow secrets, indicating a path from a public issue to sensitive CI data. The exact nature of the flaws, such as whether they involve prompt injection or improper handling of issue content, is not detailed in the provided text. The consequence, as stated, is that CI workflow secrets can be accessed, which typically include credentials and tokens used in automated builds and deployments. The article does not mention any patches or vendor responses, so no remediation information is available. The report highlights a security risk for developers using these AI coding assistants in environments connected to CI/CD systems.
CI workflow secrets can be exposed via GitHub issues, risking credential theft.