Confirm the query's target and shape
First verify which managed database, collection, or table the operation addresses. Similar names can make it easy to inspect one place while the records live in another. Compare the field names in the query with the structure available to the agent, including capitalization and nested paths where relevant.
Review each filter condition and its value. A status may be stored with different spelling or capitalization; a date may include a time component; a numeric value may not match a text value that looks similar. Check whether an empty filter is intentional and whether multiple conditions are combined more narrowly than you meant.
Use a small, read-only diagnostic
Start with a simple query that checks whether the collection or table has records, then add one condition at a time. If a broad query returns records but a filtered one does not, the issue is likely in a field or value condition. Keep the diagnostic read-only and avoid changing data merely to make a test query return something.
When asking an agent to investigate, name the target collection or table and describe the business meaning of the values you expect. Review the query and a sample result before using it in a workflow or decision. This makes it easier to distinguish a genuinely empty dataset from a mismatch in the query's assumptions.