Summary

Extend read_file tool to support reading office documents (DOCX, XLSX, PPTX) by connecting to the existing extract_text() utility in nanobot/utils/document.py.

Context

Wire up the existing office document extractors in document.py to
ReadFileTool by adding an extension guard and _read_office_doc() method
that follows the established PDF pattern. Handles missing libraries,
corrupt files, empty documents, and 128K truncation consistently.

Intentionally minimal: does not refactor document.py error protocol or
unify truncation footers to avoid behavioral changes in existing PDF path.

Why

The ReadFileTool supported UTF-8 text, images, and PDFs, but rejected .docx, .xlsx, and .pptx files with "Error: Cannot read binary file". Meanwhile, nanobot/utils/document.py already contained complete extraction logic (_extract_docx(), _extract_xlsx(), _extract_pptx()) with proper library guards, error handling, and truncation — it was simply not wired up to the tool layer.

Solve

Connected the two layers by adding an extension guard in execute() (matching the existing PDF pattern) and a new _read_office_doc() method that delegates to document.extract_text().

Changes

Tests

Completed (29 passed, 2 skipped — pre-existing PDF skips):

• DOCX returns extracted text
• XLSX returns extracted text with sheet headers
• PPTX returns extracted text with slide headers
• Missing library returns actionable error message
• Corrupt file returns error (not unhandled exception)
• Unsupported extension returns error
• Empty document returns descriptive message (consistent with PDF)
• Large document (>128K) is truncated with footer
• Small document is not truncated
• Error responses are not truncated
• Tool description mentions document support
• Tool description no longer says "cannot read"
• All existing tests (text, PDF, image, dedup, device blacklist) continue to pass

Not yet done:

• Boundary test at exactly _MAX_CHARS (off-by-one guard)
• Error path tests for .xlsx and .pptx (currently only .docx tested)
• Integration test with real office document libraries (current tests mock extract_text)

Enhancement Directions

1. Pagination for office documents — PDF has a pages parameter; office docs lack equivalent. Large XLSX with many sheets truncates at 128K with no way to continue. Consider adding sheet/slide-level pagination in a follow-up.

2. Structured error protocol — Error detection relies on startswith("[error:") string sniffing. A shared constant or discriminated return type from extract_text() would eliminate the fragile coupling between filesystem.py and document.py.

3. Dead code cleanup — The result is None guard in _read_office_doc() is unreachable because the extension gate already filters to .docx/.xlsx/.pptx. Harmless but misleading to future maintainers.