Skip to content

Security: beyhangl/evalcraft

Security

SECURITY.md

Security Policy

Supported versions

Security fixes land in the latest release on PyPI. Older versions are not patched; upgrade with pip install -U evalcraft.

Reporting a vulnerability

Please report vulnerabilities privately through GitHub's private vulnerability reporting rather than a public issue. If that page is unavailable, open an issue that only says you have a security report to share, without any details, and a private channel will be arranged. Include the evalcraft version, what you did, and what happened. You should get a first response within a week.

What to keep in mind

  • Cassettes can contain secrets and personal data. They record what your agent sent and received. Review a cassette before committing it, and use evalcraft sanitize (or CaptureContext(redact=True)) to redact known secret and PII patterns. Redaction is pattern-based and not a guarantee.
  • Replay never calls a model, but the live scorers (LLM-as-judge, RAG, pairwise) and live-eval do, and they send cassette content to the model provider you configure.

There aren't any published security advisories