Skip to content

Security: mblauberg/provenant

Security

SECURITY.md

Security policy

Scope

Security reports may cover Provenant's harness scripts, skill instructions, model routing, receipt validators and bundled third-party components on the default branch. This project orchestrates tools that may execute commands or access local files; an instruction-level authority bypass is in scope even when it is not conventional application code.

Reporting

Use GitHub's private vulnerability-reporting or draft security-advisory flow for this repository. Do not put credentials, private prompts, personal data or exploit details in a public issue. If private reporting is unavailable, open a content-free issue asking the maintainers to enable a private channel.

Include the affected revision and path, expected authority boundary, minimal reproduction, impact and any safe mitigation. Remove real secrets and personal data from examples.

Safe use

  • Inspect scripts and model routes before running them on a trusted machine.
  • Grant the narrowest filesystem, credential, network and deployment authority required for the task.
  • Treat model output, web content, repository text and third-party agent output as untrusted input rather than authority.
  • Keep credentials in the platform's secret store or environment, never in skills, prompts, receipts, logs or committed configuration.
  • Review third-party licences and code before enabling optional components.

If a secret is exposed, revoke or rotate it first, then remove it from the working tree and every published Git object. Deleting the current file alone does not remove it from repository history.

There aren't any published security advisories