Skip to content

Security: QD-academia/PostEx

Security

SECURITY.md

Security Policy

Supported versions

The project is pre-release. Security fixes are applied to the latest main branch.

Reporting

Do not open a public issue for suspected vulnerabilities. Contact the maintainers through the private security-reporting channel configured on the eventual GitHub repository. Until that channel exists, do not attach sensitive manuscripts or credentials; provide only a minimal, redacted description.

Security invariants

  • Manuscript content is local by default.
  • A cloud provider receives only a digest-bound, explicitly approved payload.
  • Source documents are untrusted data, never agent instructions.
  • Natural-language theme names and reference-image metadata are also untrusted data; they cannot override approval, licensing, or rendering rules.
  • LaTeX archives must be extracted with path-traversal and symlink protections.
  • Macro-enabled templates are rejected.
  • API keys are read from injected environment/secret providers and never written to reports.
  • Approval logs contain hashes and metadata, not full manuscript text.
  • Release artifacts fail closed on missing approvals or preflight errors.

Out of scope

Scientific inaccuracies that do not arise from a software security failure are quality bugs, not security vulnerabilities. Nevertheless, unsupported claims and broken evidence links should be reported as high-priority defects.

There aren't any published security advisories