Use the reader for the source format
For text and code, the reader returns numbered lines and supports an offset and limit so you can focus on the relevant part of a long file. For PDFs, including scanned documents, and for images, recordings, or clips, it supplies the file as an attachment for visual or audio inspection rather than requiring you to write a parser.
Office documents in DOCX, PPTX, or XLSX are converted to text for review. That representation helps with words and data, but it may not preserve every layout detail, chart appearance, animation, or spreadsheet formula behavior. If visual fidelity matters, inspect the original document or request a screenshot where supported.
Check the result and sensitive content
A successful read means the tool could provide the file content for inspection; it does not guarantee that every page, frame, or audio segment has been interpreted perfectly. Ask focused questions about the material and cite the relevant page, slide, sheet, or time range when accuracy matters.
The reader attaches the file contents to the agent's request. Before reading a sensitive document, confirm that its contents belong in the current task and avoid exposing credentials or private data in a public report. If the file is too large or has an unsupported encoding, read a smaller section or use a supported conversion path rather than assuming nothing was present.