Skip to content

Security: Vishisht16/Humane-Proxy

.github/SECURITY.md

Security Policy

Supported Versions

v0.5.6 is the first stable release; everything before it was published as a pre-release and is not supported.

Version Supported
0.6.x Yes
0.5.6 - 0.5.7 Critical fixes only
< 0.5.6 No (pre-releases)

Reporting a Vulnerability

If you discover a security vulnerability in HumaneProxy, please do not open a public issue.

Instead, report it through one of these channels:

  1. GitHub Private Vulnerability Reporting: Navigate to the Security tab of this repository and click "Report a vulnerability".

  2. Email: Send details to mishra@vishisht.tech with the subject line [SECURITY] HumaneProxy vulnerability report.

What to Include

  • A description of the vulnerability and its potential impact.
  • Steps to reproduce (proof of concept if possible).
  • The affected version(s).
  • Any suggested fix or mitigation.

What to Expect

  • Acknowledgement within 48 hours.
  • A fix timeline communicated within 7 days.
  • Credit in the release notes (unless you prefer to remain anonymous).

Scope

The following are in-scope for security reports:

  • Bypasses of the safety pipeline (e.g., a message that should be flagged as self_harm but is classified as safe).
  • Injection attacks via webhook payloads, admin API, or MCP server.
  • Denial-of-service vectors (e.g., ReDoS in keyword matching).
  • Information disclosure through the admin API or escalation logs.

Out of Scope

  • Social engineering of LLM providers (OpenAI, Groq) — these are upstream services.
  • Vulnerabilities in dependencies that do not directly affect HumaneProxy's behaviour.
  • Theoretical attacks that require physical access to the host machine.

Disclosure Policy

We follow coordinated disclosure. Please allow us a reasonable window (typically 30 days) to address the vulnerability before public disclosure. We will work with you on a timeline.

Logging & Data-Exposure Hardening

As of v0.5.5:

  • Webhook logs contain only the scheme and host of webhook URLs — never the path, query, or fragment, which carry routing tokens for Slack/Discord/Teams.
  • Webhook error logs record HTTP status and response length only; response bodies are logged exclusively at DEBUG level.
  • The reverse proxy never returns raw upstream response bodies to clients; non-JSON upstream responses produce a generic error message, with the body available to operators at DEBUG level.

There aren't any published security advisories