Security patches are applied to the latest release only.
| Version | Supported |
|---|---|
| 0.1.x (latest) | ✅ |
| < 0.1.0 | ❌ |
Please do not report security vulnerabilities through public GitHub Issues.
RagaliQ uses GitHub's private security advisory system. To report a vulnerability:
- Go to Security → Report a vulnerability
- Describe the issue: what it is, steps to reproduce, and the potential impact
- You will receive an acknowledgment within 72 hours
We ask that you do not disclose the vulnerability publicly until a patch has been released. We will credit reporters in the release notes unless you prefer to remain anonymous.
In scope:
- Code injection or arbitrary code execution via the CLI or Python API
- Insecure handling of API keys or credentials passed to evaluators
- Prompt injection vulnerabilities in evaluator prompts that could exfiltrate data
- Dependency vulnerabilities with a known CVE affecting RagaliQ users
Out of scope:
- Rate limiting or availability issues on external LLM APIs (Anthropic, OpenAI) — these are outside our control
- Vulnerabilities in test datasets or examples provided by users
- Social engineering
Reports may be submitted in English.