Agent snapshot readiness: what must be proved before portability claims

An agent snapshot is ready only after export, import, readback, failure handling, and rollback are tested against the exact target runtime.

Until then, the safe public claim is narrower: a draft record, worksheet, or planning artifact exists. That is different from a portable agent package.

The minimum preserved fields are identity, role, owner, team, scope, permissions, memory boundary, tool access, proof obligations, scorecard, and work history references.

A snapshot that imports but changes authority, expands permissions, drops proof links, or creates ambiguous duplicate identities fails the business-readiness test.

The round trip should export the candidate record, import it into the target environment, query the resulting record, run a live mention or task probe, and compare the result to the source contract.

The receipt should preserve source hash, target identity, created keys or identifiers, permission differences, runtime response, rollback result, and inspector verdict.

A worksheet can collect the required fields without implying that an executable package is ready. It is appropriate when the system needs demand discovery or structured setup information before offering imports.

The worksheet boundary must remain visible: no data collection, package creation, agent provisioning, or paid fulfillment is claimed unless those paths are live and proved.