Skip to content

Security: Clyra-AI/axym

Security

SECURITY.md

Security Policy

Axym is designed for deterministic, local-first evidence handling. Security bugs that could weaken proof integrity, schema validation, output safety, release verification, or secret handling should be reported privately before public disclosure.

Private reporting path

  • Use GitHub Security Advisories as the private reporting path for vulnerability reports in this repository.

Fallback public path

  • If GitHub Security Advisories are unavailable, open a minimal public GitHub issue without exploit details and ask maintainers to continue through the path documented here.

Supported security boundaries

Please report issues involving:

  • proof record signing or chain verification bypass
  • bundle integrity or provenance verification gaps
  • fail-open behavior where Axym should fail closed
  • schema validation bypass or unsafe artifact acceptance
  • secret leakage, raw evidence exfiltration, or unsafe output path handling
  • release artifact, checksum, signature, or SBOM verification issues

How to report

  • Prefer GitHub Security Advisories before opening a public issue.
  • If GitHub Security Advisories are unavailable, use the fallback public path above and avoid posting exploit details, raw secrets, or sensitive evidence.

What to include

  • affected version or commit
  • installation path used
  • exact command or workflow involved
  • minimal reproduction steps
  • expected vs actual behavior
  • impact on determinism, integrity, privacy, or policy enforcement

Disclosure expectations

  • Give maintainers a reasonable window to reproduce, patch, and publish guidance before full public disclosure.
  • Avoid publishing raw exploit material that would expose real user evidence or secrets.

Release verification

When validating a release-related report, include the verification path used, for example:

./scripts/release_go_nogo.sh --dist-dir dist --binary-name axym

and any checksum, signature, or provenance output relevant to the issue.

There aren’t any published security advisories