Skip to content

Security: johnzastrow/niva

SECURITY.md

Security Policy

Reporting a vulnerability

Please report security issues privately — do not open a public issue for an unfixed vulnerability.

We aim to acknowledge a report within a few days and will coordinate a fix and disclosure timeline with you.

Supported versions

niva is pre-1.0 and ships fixes on the latest minor release. Please reproduce on the most recent release before reporting.

Security posture (what reduces risk by design)

  • Zero runtime dependencies — niva imports only the Python standard library and the QGIS APIs already present in your QGIS install. There is no third-party package tree to compromise.
  • No dynamic code execution of input — niva never eval/exec/pickles a flow or a file. The import command parses foreign PyQGIS scripts with ast.parse (static analysis only; the script is never executed).
  • Credentials stay in QGIS / the environment — database and notification secrets are read from the QGIS connection store or environment variables, never from flow text, and are never written to logs or the run journal.
  • Network access is constrained — remote show fetches only http(s) URLs (scheme allowlist), validates TLS, and refuses XML carrying a <!DOCTYPE> (XXE guard).

For contributors

  • Commits to main go through pull requests with required CI; force-pushes and branch deletion are disallowed.
  • GitHub Actions are pinned to commit SHAs and updated via Dependabot.
  • Never commit secrets. Run the staged-content scan in ~/.claude/CLAUDE.md / rely on GitHub push protection before pushing.

There aren't any published security advisories