Skip to content

Security: Open-Agent-Tools/Postcard

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Please do not open a public GitHub issue for security reports.

Instead, use GitHub's private vulnerability disclosure: https://github.com/Open-Agent-Tools/Postcard/security/advisories/new

We'll acknowledge within a few days and work with you on a fix and coordinated disclosure timeline.

Scope

oat-postcard runs entirely on the local filesystem. The threat model assumes any user account with read access to ~/.oat-postcard/ is trusted. In particular:

  • Postcards are plain JSON on disk, readable by any process running as the user.
  • The ledger is a local git repo; anyone who can read it can replay all history.
  • No authentication or encryption between sessions — sessions on the same machine trust each other.

Issues we're interested in:

  • Path-traversal or injection in address/title/body fields.
  • Atomicity bugs that could corrupt the ledger.
  • Privilege escalation from hook scripts.

Issues we'd treat as feature requests, not security bugs:

  • Adding end-to-end encryption.
  • Network / cross-machine transport.
  • Multi-user isolation.

Supported versions

Only the latest 0.x release receives fixes until a 1.0 is tagged.

There aren’t any published security advisories