AgentOS is the vendored development/review framework this repository is built with. It is not part of the APIx statistical product and produces none of APIx's published numbers. Its version numbers, release notes and readiness statements describe AgentOS, never APIx.
For APIx — the airfare price index — see README.md and docs/capability-matrix.md.
See SUPPORTED_VERSIONS.md for which versions of AgentOS receive security updates.
AgentOS is typically deployed as a private internal engineering template. However, if you discover a security issue — particularly in the validation scripts, bootstrap tooling, or runtime kernel — please follow responsible disclosure.
- Do not open a public GitHub issue for security vulnerabilities.
- Email the maintainer directly or use the repository's private security advisory feature.
- Include:
- Description of the vulnerability
- Steps to reproduce
- Potential impact
- Suggested fix (if known)
- Acknowledgement: Within 48 hours
- Initial assessment: Within 5 business days
- Fix timeline: Depends on severity — critical issues within 7 days
AgentOS follows these security principles by design:
- No network calls in core framework — All runtime scripts operate locally
- No credentials in repository —
PROJECT_CONFIG.yamlcontains no secrets - No execution of untrusted code — Validation scripts are read-only analyzers
- Vendor isolation — AI provider credentials stay in
integrations/only - Policy-driven access — All routing decisions are governed by YAML policies
Security issues in these areas are in scope:
| Area | Risk Level |
|---|---|
tools/scripts/ execution |
High — runs Python with filesystem access |
runtime/ kernel modules |
Medium — state machine and scheduler logic |
.agentos/config.yml |
Medium — routing configuration |
integrations/ adapters |
High — AI provider credentials may be present |
validation/runner/ |
Low — read-only analysis |
For security concerns, contact the project maintainer through your organization's internal channels.