Skip to content

Security: Drunklvu/hermes-codex-bridge

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Do not open a public issue for security vulnerabilities.

Please report privately instead — this project is small and maintained by one person, so email is the fastest path:

  • Email: bridge-security@users.noreply.github.com (maintainer's GitHub noreply address — the exact mailbox is resolved at the maintainer's side; if this address bounces, open a private issue via the GitHub security advisory flow: repo → Security → Report a vulnerability)

You should receive an acknowledgment within 3 business days. If you do not hear back, escalate via a private issue on the repository (GitHub's confidential reporting).

What to include

  • Affected version / commit SHA
  • Steps to reproduce (minimal)
  • Impact description
  • Suggested fix (optional)

Security posture of this project

  • Loopback-only by default — the bridge and sidecar bind 127.0.0.1; non-loopback binding is refused unless --require-token is set and a dedicated sidecar token is provided (the bridge admin token is never reusable for remote access).
  • Zero-dependency corecodex_a2a_bridge.py / hermes_mcp_server.py import stdlib only, reducing the supply-chain surface.
  • Secrets scan in CI — every push is scanned for credential patterns, local paths and business-specific words (tracked files + git history).
  • Token handling — runtime tokens (bridge.token, inbound.token) are gitignored, generated on first run, and never logged.

Known trade-offs (documented, not silently ignored)

  • AgentCard signing is fail-open: a --sign-key failure logs and continues serving an unsigned AgentCard (discovery endpoints are anonymous per the A2A spec). If you require strict verification, gate clients accordingly.
  • /health and AgentCard stay anonymous in remote mode — they are discovery endpoints; they expose no task data.

Supported versions

Only the latest main is actively maintained. There are no LTS releases.

There aren't any published security advisories