Skip to content

Security: divmora/code-review-bot

SECURITY.md

Security Policy

Supported versions

Security fixes go into the latest minor release.

Reporting a vulnerability

Please do not open a public issue. Report privately:

Include a description, impact, reproduction steps and any suggested fix. We acknowledge within 48 hours, triage within 5 business days, and coordinate disclosure with you.

Operating the bot safely

  • What leaves your network: changed files, their diffs and (for files up to full_file_max_lines) their full content are sent to the configured model provider. Pick a provider and data-retention terms that fit your code's sensitivity, or run a local model (Ollama, vLLM).
  • Least-privilege tokens:
    • GitHub token in Actions: contents: read, pull-requests: write.
    • GitLab: a dedicated bot user or project/group access token with api scope, Developer role.
    • Bitbucket Cloud: a repository access token with Pull requests: Write and Repositories: Read (add Pipelines/build status write for --commit-status).
  • Untrusted PRs: PR content is passed to the model as data, and repo review rules are read from the base commit. Treat model output as advisory: the CLI only comments and never approves, merges or pushes.
  • Forks: don't expose provider keys to workflows triggered by forked PRs (pull_request_target with checkout of fork code is unsafe). The example workflow skips forks.
  • TLS: skip_tls_verify exists only for self-managed GitLab with private CAs; prefer adding the CA to the image trust store.
  • Secrets in config: use ${ENV_VAR} references or api_key_env instead of literal keys, and keep codereview.yaml out of git (it is in .gitignore).

There aren't any published security advisories