Mixed document batches for the assessor's office need metadata before enrichment. Which tool_choice pattern fits?
Select an answer to reveal the explanation.
Short Explanation
Don't decorate the folder before you label it. Force extract_metadata first, then run enrichment on mixed assessor batches.
Full Explanation
Mixed document batches for the assessor's office need reliable metadata before enrichment tools run. Forcing the extract_metadata tool first—via tool_choice of that named tool or an ordered agent step—establishes document type, jurisdiction, and identifiers that later enrichment depends on. Ordered tool_choice prevents enrichment from running blind on unclassified packets.
Letting enrichment tools run first with no metadata extraction step fails because enrichment lacks the keys and type context required for correct assessor workflows. Skipping metadata and enriching whatever prose Claude invents fails data governance: invented metadata is not an extract. Calling enrichment only and hoping metadata appears as a side effect fails by concept; side effects are not a contract and will not populate required fields consistently.
Exam caveat: forcing extract_metadata first fixes ordering; enrichment tools must still validate that metadata fields meet minimum completeness before mutating records. Operational check: in the assessor pipeline, set the first tool_choice to extract_metadata, assert metadata presence, then allow enrichment tools—fail the batch if enrichment is attempted without metadata.