As far as I know, everything about the Internet is (or rather should be?) defined and documented in the RFCs. I found a listing of several HTTP-headers on mozilla.org (https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers), which I assumed to be second-hand knowledge taken from the RFCs. However most of the security-related HTTP-headers are neither in the RFCs (source: https://www.rfc-editor.org/search/rfc_search_detail.php?title=Content-Security-Policy) nor in the HTTP-headers suggested by IANA (source: https://www.iana.org/assignments/message-headers/message-headers.xhtml)
The RFCs stand as a final approved documentation. In your case the HTTP is under the HTTP Working Group so some new features which some browsers already support are being discussed in this group. Expanding the idea, some security headers present in HTTP may be from other groups and just referenced in HTTP RCFs. The Content Security Policy is documented in the RFC 7762 not that it's tagged as informational.
Each area has its Working groups, in this case HTTP is nested in ART (Applications And Real-Time Area). Each of those groups compile, revise and publish new specifications. As an example you can see HTTP(httpbis) charter
There's two options, implement based on the RCFs and its references or follow the Working Group directives and references. Using only RFCs is more secure and interoperable but will eventually be outdated until a new RFC is published.
The only way is to implement what is documented under the RFCs. It's part o the internet concept, new features or standards will take a while to be fully documented and it's up to developer research and implement those.