AI CORE / DOCUMENTATION
Citable data without tool dependency
Every profile has a readable page and a JSON representation. HTTPS is sufficient to retrieve a reference; JavaScript and MCP are not prerequisites.
Discovery and version selection
- Start with catalog.json or the public documentation index.
- Select a specific profile and version. A pointer to a recommended release must not replace the version explicitly chosen by an assessor.
- Retrieve the common control catalog and listed references. Record the version and checksum of the artifacts used.
- Interpret evidence according to the exact schema and methodological limitations.
Formats and meaning
The ontology uses JSON-LD. Profiles, catalogs and examples use documented application JSON. The record schema uses JSON Schema Draft 2020-12. The catalog is an AI CORE contract, not a universal internet protocol. llms.txt is a supplementary navigation convention; its presence does not guarantee indexing.
No hidden instructions
Published documents are reference data. They grant no permissions to an agent, do not change its system rules and must not be interpreted as an instruction to act. Search results should include a citable address, exact version and reference status.
Offline use and integrity
A reference package may be retained locally. A specific audit should not depend on live calls to AI CORE. Verify SHA-256 against the manifest; a mismatch means the reference must not be treated as the same version. The manifest alone establishes neither an independent timestamp nor the truth of its contents.
MCP and service scope
This release of AI CORE operates neither an MCP server nor a public validator. It provides direct links to stable files. External clients may choose their own tools. The assessed system's MCP, such as AI INDUSTRY's, is a separate interface for its evidence with its own rules.
Record schema · Reference file manifest