Skip to content

Security: evaleval/auto-benchmarkcard

SECURITY.md

Security

Auto-BenchmarkCard is research code, not a hardened public service. The pipeline follows candidate URLs from its inputs and fetches papers, repository pages, dataset pages and other web sources. It does not provide a complete blocklist for localhost, private networks, cloud metadata endpoints or other sensitive destinations.

Do not expose the CLI or the webhook to untrusted inputs without adding URL validation, network egress controls, authentication, request limits and logging appropriate to the deployment. Run jobs in an isolated environment with least-privilege credentials. Never place credentials in committed files; use environment variables or a secret manager.

Report suspected vulnerabilities through GitHub's private vulnerability reporting for this repository rather than in a public issue, and do not include live credentials or sensitive source material in the report.

There aren't any published security advisories