ZenoXCare · Security
Coordinated vulnerability disclosure
ZenoXCare accepts security reports from anyone — researchers, partners, customers, and AI agents alike. This page is the canonical policy. It is also published as a machine-readable manifest at /.well-known/security.txt per IETF RFC 9116, and surfaced inside the A2A agent card under policies.disclosurePolicy.
How to report
Send a written description of the issue, reproduction steps, observed impact, and (optionally) a proof-of-concept to one of the contacts below. Encrypted submissions are welcome but not required.
- RFC 9116 manifest
- /.well-known/security.txt
- A2A agent policies
- /.well-known/agent.json
- Acknowledgments
- /security/hall-of-fame
Response timeline
- Acknowledgement≤ 2 business days from receipt
- Initial triage + severity≤ 5 business days from confirmation
- Coordinated disclosure target90 days from confirmation
- ExtensionMutually agreed for high-complexity issues; we will tell you why and when in writing
In scope
- *.zenoxcare.comAll web hosts under the ZenoXCare apex.
- Mobile applicationsApps published by ZenoXCare.
- /api/public/*, /api/v1/*, /.well-known/*Public API surface (REST + A2A + machine-readable discovery).
- /network/sandboxVerification Network developer sandbox.
- /.well-known/agent.json + /api/public/a2a/*A2A agent surface and dispatch endpoints.
Out of scope
- Third-party services we depend on (Paystack, Twilio, Resend, Vercel, Sentry) — please report directly to those providers.
- Denial-of-service tests against production (use the sandbox + your own API key + your own tenant).
- Social engineering of staff or partners.
- Findings that require physical access to user devices.
- Reports based purely on outdated browsers / OS versions without an exploitable path.
Safe harbor
- We will not pursue legal action against researchers acting in good faith, staying within scope, and following coordinated disclosure.
- Please use the dedicated `security-research@zenoxcare.com` test tenant when possible.
- If a finding requires accessing data that is not your own, stop, document the impact, and report immediately — do not exfiltrate or persist data beyond what is necessary to demonstrate the issue.
Why this exists for AI agents too
Peer agents that consume ZenoXCare's A2A surface (/.well-known/agent.json) can read these same disclosures programmatically through the policies block. Our prompt- and tool-injection defense posture is declared as tier-0 there — the runtime guardrail pipeline (lib/ai/agent-guardrail-pipeline.ts) runs on every A2A dispatch + every internal AI call. If you discover a way to bypass it, that's a security issue under this policy and we want to hear about it.