Skip to content

Security: hoid-labs/hoid-core

Security

SECURITY.md

Security Policy

Supported versions

The latest minor release of hoid receives security updates. Earlier minor versions are not patched; please upgrade.

Version Supported
Latest minor (0.x where x is the highest released) yes
All earlier versions no

Reporting a vulnerability

Do not file a public GitHub issue for security vulnerabilities. Public disclosure before a fix is available makes it easier for attackers to exploit the issue.

Email a description of the vulnerability to <security-contact> (replace with the address published on the project page). Include:

  • A description of the vulnerability and its impact
  • A minimal proof-of-concept or reproduction steps
  • The version of hoid affected
  • Your environment (Python version, OS, optional extras in use)

You can expect an acknowledgment within 3 business days. If you do not hear back, follow up on the same thread.

Disclosure timeline

  • Day 0 — you report the issue. We acknowledge within 3 business days.
  • Day 0–14 — we investigate, develop a fix, and prepare a release. We may request additional time if the fix is non-trivial.
  • Day 14–90 — coordinated disclosure. We share the fix with you under embargo and agree on a public disclosure date. We aim to publish the fix within 90 days of the initial report.
  • After public disclosure — the security advisory is published on GitHub, the fix is released in a patch version, and the public changelog credits you (if you wish).

We will not pursue legal action against researchers who make a good-faith effort to follow this process.

Scope

The following are in scope:

  • Code execution, sandbox escape, or arbitrary file/network access from a tool or extension where the user did not opt in
  • Credential or secret disclosure through default configuration, logging, or error messages
  • Bypasses of AuthGate, AuthContext, or any other access-control mechanism
  • Insecure defaults shipped by the library itself (not by user code)
  • Vulnerabilities in the agent loop, MCP transport, OIDC flow, or RAG store

The following are out of scope:

  • Vulnerabilities in user code, user-defined tools, or user-defined guardrails
  • Issues in optional third-party packages (file those upstream)
  • Denial of service via malicious prompts (the model is the user; rate-limit at the deployment layer)

Security best practices for users

  • Pin hoid to a specific minor version in production (hoid>=0.x,<0.(x+1))
  • Run agents with the smallest possible tool surface — disable any tool the agent does not need
  • Set CA_BUNDLE_PATH if your LLM endpoint uses a private CA
  • Use AuthGate to enforce per-user tool access when serving agents to multiple users
  • Subscribe to GitHub security advisories (Watch → Custom → Security alerts) to be notified of new disclosures

Contact

Email: <security-contact> (replace with the address published on the project page)

There aren't any published security advisories