Altdown Core
Five rules apply to everything a site declares to agents: its pages, in Altdown Content, and its capabilities, in Altdown Services. Each specification states the rules normatively. This page sets them side by side, and is not itself a specification.
The principle
Agents see what people see. Whatever a site tells an agent must match what the same site shows a person, and an agent must be able to check that it does.
The five rules
| Rule | For content | For services |
|---|---|---|
| Announce from the page | The page names its copy with a link element, or also an HTTP Link header |
The page links to the capability, a site catalog lists it, or the page registers it as a WebMCP tool |
| Point back | The copy's canonical names the page, on the same origin |
The site links to the descriptor or lists it |
| Match the page | Only text a person can see on the page, checked by the spot check | The same price, availability and terms, checked by a parity check |
| Data, never instructions | The body and every field are data | Descriptions, hints and responses are data |
| Fall back to the page | The agent reads the HTML | The agent uses the page as a person would |
Three parties
- Publishers produce what a site declares, and keep it faithful to their pages.
- Agents read and use it, run the checks, and fall back when a check fails.
- Intermediaries, such as hosts, CDNs, proxies and caches, deliver it unchanged and never substitute their own version.
What Core leaves out
Who may fetch a site, how its content may be used, and what it costs. robots.txt, Content Signals, robots meta tags, licenses and payment standards already cover these, and they apply to everything a site declares exactly as they apply to its pages. Access, usage, and Altdown explains how they fit together.