Skip to content

Security: h1ddenpr0cess20/agent-smithers

Security

SECURITY.md

Security Policy

We take security seriously. This document explains how to report vulnerabilities and provides guidance on hardening deployments.

Reporting a Vulnerability

  • Prefer private disclosure. If GitHub Security Advisories are available for this repository, open a private advisory.
  • If not available, open a minimal public issue asking for a secure contact channel (do not include sensitive details), or DM the maintainer via project channels.
  • Provide:
    • A clear description of the issue and security impact
    • Minimal reproduction steps and affected version/commit
    • Environment details (OS, Python version)
    • Proof‑of‑concept if safe to share privately

We aim to acknowledge reports within 7 days and, where applicable, provide a remediation plan or patch within 30 days. Timelines may vary based on complexity and severity.

Scope

In scope:

  • Authentication and device verification flows
  • E2EE assumptions and key handling insofar as the bot interacts with them
  • Injection risks from user content (e.g., Markdown/HTML rendering, replies)
  • Command handlers and admin‑only gating
  • Configuration secrets and sensitive data handling
  • Network transport and dependency vulnerabilities that affect the bot

Out of scope:

  • Vulnerabilities in Matrix homeservers, LLM providers, or other third‑party libraries/services
  • Social engineering or policy issues unrelated to code or configuration
  • Unsupported branches or old releases
  • Misconfigurations in self‑hosted environments not caused by code defects

Coordinated Disclosure and Safe Harbor

Please make a good‑faith effort to avoid privacy breaches, service disruption, or data loss. We support common‑sense safe‑harbor principles for security research conducted and reported responsibly and privately.

Hardening Guide

See the Security guide: docs/security.md for practical steps to harden your deployment (device verification, room controls, config secrets, logging, and more).

There aren't any published security advisories