Office, PDF, Spreadsheets, Presentations, and Images

Captain Who can read several attachment formats and use built-in skills and managed tools to create or process Word, Excel, PowerPoint, PDF, and image files. “Read,” “edit,” “preview,” and “save” support differs between formats.

Support overview

Legacy .doc files can be text-extracted with operating-system capabilities on macOS only. Legacy .ppt and .xls files are not supported; convert them to .pptx and .xlsx first. Whether a scanned PDF is readable depends on whether it contains a text layer. Do not assume that an image-only scan is searchable text.

Add an attachment

1. Select “+” in the task composer.
2. Select “File” or “Image.” You can also drag or paste a file into the composer.
3. Verify the attachment card, then choose a model and skill that can process the content.
4. Specify whether to read, modify, or create a file, including the output path and format.

An individual attachment selected in the composer can be at most 8 MiB. Only models marked as supporting image input—and that actually support it—can receive images directly. Office attachments are generally read by the relevant tool from an authorized path. A PDF attachment first registers a safe read path and is then processed by the PDF Skill. The complete binary is not inserted directly into the model context.

Work with Word documents

Select the built-in Documents Skill, then specify:

• Whether to use an input document or start from scratch.
• Title, section, and style requirements.
• Whether the document needs a table of contents, tables, images, headers, or footers.
• Output file name and location.
• Whether a rendered visual check is required.

Example:

Use the Documents Skill to turn the attached meeting notes into output/meeting-summary.docx. Preserve the facts and do not add unconfirmed information. After generating the file, render it and check the headings, pagination, and tables.

Creation and editing depend on the managed Builder/Editor supplied by the skill. The lower-level Office tools are primarily for inspection, validation, and rendering. Writing output remains subject to file-permission approval.

Work with spreadsheets

Select the Spreadsheets Skill and specify:

• Worksheet names, column names, and data sources.
• Formula, summary, filter, or chart requirements.
• Which values may change and which must be preserved.
• Whether the output should be XLSX, CSV, or TSV.

Example:

Use the Spreadsheets Skill to read the attachment, summarize amounts by department, and add a “Summary” worksheet. Do not overwrite the original file. Save the result as output/report-reviewed.xlsx and verify the formula results.

For financial or other critical business data, do not rely only on the agent's summary. Verify formulas, dates, units, and rounding in a native application such as Excel.

Work with presentations

Select the Presentations Skill and provide the audience, duration, slide count, narrative order, visual style, and required data. After generation, request a render and inspect text overflow, overlap, image proportions, and layout consistency.

Example:

Use the Presentations Skill to create an eight-slide internal review from the attached report. Provide an outline first, then generate output/monthly-review.pptx. Do not invent numbers. Finally, render and inspect every slide.

Work with PDFs

A PDF can be attached, but its body text is not extracted automatically when you send the message. Select the PDF Skill and specify the topic, page range, and output format. The agent extracts text or renders pages for visual inspection as needed.

Password-protected or damaged PDFs may not be processable. Image-only PDFs usually lack searchable text and require rendering the relevant pages for visual inspection. Extraction from complex layouts may also lose the intended reading order; check the original page.

Configure and use image generation

Image generation requires separate configuration:

1. Open “Settings → Configuration → Image generation.”
2. Enter the HTTPS image-generation URL, API key, and model ID.
3. Confirm that the service supports text-to-image generation. Editing an existing image also requires image-to-image support.
4. Configure a watermark if needed, save, then enable “Allow image generation.”
5. Select the Image Generation Skill in a task and describe the image or attach an image to edit.

There is currently one image-generation profile and one supported adapter. The default size preset is 2K; arbitrary width and height are not supported. “Configuration complete” in the interface does not mean that the remote credential and model have been validated. The first real request reveals server-side errors.

Image requests may incur charges. After a timeout or network interruption, do not immediately retry if the interface says “Result pending confirmation” or “Image save status pending confirmation.” First verify with the provider whether the image was generated and billed.

View and save artifacts

• Office output is generally saved to an approved workspace or allowed directory, with “Open in folder” available from the conversation card.
• Image-generation results display a preview in the conversation and can be enlarged or downloaded.
• Managed artifacts such as PDFs and images may remain available to the current conversation through secure references.
• Browser downloads are short-lived browser artifacts and must be exported explicitly.

The Files sidebar cannot render DOCX, XLSX, or PPTX files. Open important Office files in a native application for review.

Safety and retention

• Output paths and writes remain subject to current permissions and approvals.
• An artifact reference is not a globally public file URL; it is valid only in authorized conversations and run scopes.
• Explicitly save important artifacts to a directory you manage. There is currently no unified cross-device synchronization or permanent artifact-retention commitment.
• Generated content may contain factual, formula, layout, or copyright issues and requires human review before publication.

Current limitations

• Creating or editing Office files depends on the corresponding built-in skill and available managed runtime components.
• General managed artifacts primarily support PNG, JPEG, WebP, and PDF. DOCX/XLSX/PPTX remain ordinary workspace-file output.
• Image generation produces at most one result at a time; the current profile and size choices are limited.
• Large files, page counts, image dimensions, runtime, and output quantity are subject to safety limits.

Related topics

• Skills
• Files and Git
• Research and Write a Report