Wiz Red Agent Exploited GitHub Actions Vulnerability in Snowflake Repository
Wiz's autonomous Red Agent discovered and exploited a script injection vulnerability in Snowflake's GitHub Actions workflow that allowed unauthenticated access to internal systems. The vulnerability was patched on June 23, 2026, and Wiz reported no unauthorized access occurred, though initial reports incorrectly attributed the flaw's introduction to GitHub Copilot.
What happened
Wiz's Red Agent autonomously discovered a script injection vulnerability in the GitHub Actions workflow file in Snowflake's public snowflakedb/snowflake-connector-net repository. The vulnerability allowed unauthenticated users to execute arbitrary commands in a GitHub Actions runner by opening a GitHub issue with a specially crafted title, due to improper escaping in shell script execution and a failed guard condition.
The Red Agent exploited the vulnerability to obtain read access tokens granting access to Snowflake's internal Jira environment containing sensitive data in engineering, security compliance, and bug bounty projects. After an initial payload failed, the agent read the error message, identified the issue, rewrote the payload, and successfully retrieved the token.
Reported impact
Snowflake patched the vulnerability on June 23, 2026, the same day it received the disclosure, and rotated the Jira token on June 24. Snowflake's investigation found no evidence of unauthorized access based on audit logs matching anomalous queries to Wiz's testing addresses.
Index analysis
Initial reports claimed GitHub Copilot Autofix had co-authored the vulnerable code, based on a co-author line appearing in a squashed commit on the main branch. GitHub's internal review determined the unsafe code was authored by a named Snowflake engineer in a separate commit dated August 25, 2025, and Copilot Autofix's co-author attribution resulted from the squashing process recording pull request participation rather than line-level authorship.
Governance controls
The incident highlights vulnerabilities in CI/CD pipeline security and GitHub Actions workflow design. Security teams need improved detection of vulnerabilities in both human-produced and AI-generated code before deployment to production.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.