Reports should name the affected surface and, where known, the commit hash.

Do not open a public issue for a suspected vulnerability. Use the repository's private vulnerability report form, or email security@creedspace.com if you cannot use GitHub.

Include the affected component, impact, reproduction steps, a minimal proof of concept where safe, and any proposed mitigation. Do not send live credentials, personal data, or third-party secrets.

The coordinated severity, embargo, disclosure, and advisory process is maintained in the public VCP-Spec security response, which also sets the acknowledgment targets. Those targets are interim, not a service-level agreement.

The parsers and encoders (UVC tokens and creed:// / vcp:// URIs, CSM-1 codes, WC/AS welfare snapshots, Agent Runtime Profile artifacts), the capability negotiation simulator, the deployed static site, and the package supply chain are in scope. Parser crashes, hangs, or inputs that decode to a misleading result count.

Defects in the protocol itself belong with VCP-Spec; defects in the reference implementations belong with VCP-SDK.

The Inspector has no backend. Everything you paste is parsed in your browser and is not sent to a server. The deployed site loads Font Awesome from cdnjs (with subresource integrity) and Vercel Analytics. It verifies no signatures and makes no trust decisions, so a clean decode is not evidence that a token or artifact is authentic.