CVE-2026-53942

Summary

A GitHub Actions workflow in conda/conda executes an external template-rendering action using attacker-controlled repository and branch references obtained from a pull request comment trigger, allowing an authenticated attacker to achieve remote code execution via unsandboxed Jinja2 template rendering with elevated privileges.

Exploitability

Exploitation requires an authenticated attacker with permission to comment on issues or pull requests in the affected repository and craft a malicious comment containing @conda-bot render or @conda-bot recreate, along with a PR-controlled repository and branch pointing to a malicious template.

Blast radius

Successful exploitation would allow an attacker to execute arbitrary code with the privileges of the GitHub Actions runner, potentially compromising secrets, infrastructure, and downstream users relying on the conda ecosystem.

Detection

Monitor GitHub Actions workflow runs triggered by issue_comment events containing @conda-bot render or @conda-bot recreate for execution of external actions (e.g., conda/actions/template-files) that perform unsandboxed template rendering; look for unusual checkout of non-default repositories or branches in the workflow run logs.

Prioritized remediation

Upgrade to commit 9c0f238 or later in the conda/conda repository, which patches the vulnerability by sandboxing the Jinja2 environment and removing unsanitized template rendering from the PR-triggered workflow path.

Analysis generated locally by garnet-coder (no data left the box). AI-assisted — verify against primary sources before acting.