Known limitations

Foundation makes shared values deterministic and portable within declared contracts. It does not prove that a downstream scientific workflow chose the right value, interpreted it correctly, or preserved it through an undeclared integration.

Contract limits

Limitation	Consequence	Honest interpretation
identifiers establish stable identity syntax, not biological equivalence	two valid identifiers may still refer to records that require domain reconciliation	use Knowledge or the scientific owner for equivalence claims
canonical JSON covers supported normalized values	arbitrary objects and environment-dependent encodings are outside the guarantee	validate inputs before claiming byte stability
fingerprints are deterministic for canonical input	they do not establish provenance, authenticity, or scientific correctness by themselves	retain source and transformation records separately
document validation enforces declared schema	a schema-valid document can still contain scientifically weak evidence	pair schema proof with the package that owns scientific meaning
migrations cover registered version paths	unregistered, lossy, or skipped versions are not implicitly compatible	require an explicit fixture and loss policy for each path
result, failure, and refusal primitives preserve outcome categories	consumers can still flatten those categories incorrectly	test the consuming branch and published representation
public test helpers enforce repository contracts	they do not substitute for package-specific scientific tests	keep owner evidence next to owner claims

Guarantee boundary

flowchart LR
    F["Foundation: shape, identity, serialization"] --> C["consumer interpretation"]
    C --> W["workflow behavior"]
    F --> P["Foundation proof"]
    C --> O["consumer proof"]
    W --> S["scientific evidence"]

Foundation proof ends at the shared contract. Claims about recommendation, execution, evidence strength, assay outcome, or biological truth require proof from Intelligence, Runtime, Knowledge, Lab, or Core as appropriate.

Reporting compatibility

State the precise envelope: for example, “canonical JSON is byte-stable for the document primitives covered by the round-trip fixtures,” or “schema version A migrates to B through the registered migration.” Avoid “serialization is stable” when only one document family or direction was tested.

When consumer evidence has not run, report the Foundation result as local contract proof and leave cross-package compatibility unresolved.