Skip to content

Security: openagentlock/OpenAgentLock

SECURITY.md

Security policy

Reporting a vulnerability

Once this repository is public, the preferred channel is GitHub's private vulnerability reporting:

(Private vulnerability reporting on private repos is gated to the GitHub Advanced Security add-on, so the form is only usable while the repository is public, or for orgs on a paid plan.)

If the form isn't reachable for any reason, open a GitHub issue with just enough detail to recognize the report as a security issue — please don't include a working exploit, reproduction steps, or PoC payloads. We'll move the conversation to a private channel from there.

We acknowledge reports within 72 hours.

Scope

In scope:

  • The CLI (cli/)
  • The control plane (control-plane/)
  • The ledger crate (ledger/)
  • The published ghcr.io/openagentlock/agentlockd Docker image
  • The published @openagentlock/cli npm package

Out of scope:

  • Misconfigurations of agent harnesses themselves (Claude Code, Cursor, Codex CLI, etc.)
  • Issues that require an attacker who already controls the host
  • Reports based on outdated dependencies without a working exploit

Disclosure

Once a fix lands and a release is cut, we will:

  1. Publish a GitHub Security Advisory with a CVE if applicable.
  2. Credit the reporter unless they request anonymity.
  3. Note the fix version in CHANGELOG.md.

Cryptographic posture

  • Ledger: Ed25519 over SHA-256 Merkle leaves, RFC 6962 odd-tail handling, RFC 8785 JCS canonicalization for signed payloads.
  • Hardware-key signer: PIV / FIDO2 via the host CLI; YubiKey not bridged into Docker (USB HID limitation).
  • TOTP signer: RFC 6238, 6-digit, default 30-second step.

If you believe any of these primitives are misused or the chosen parameters are insufficient, please file a private advisory.

There aren’t any published security advisories