Office Studio
Turn a one-line request into a finished, downloadable Office document - a real PowerPoint deck, a working Excel model, or a formatted Word report, built from genuine OpenXML with no Microsoft Office anywhere in sight. It drives OfficeCLI inside a sandboxed shell and loads its document-building skills live from GitHub, so a request becomes an actual .pptx, .xlsx, or .docx file you can open and send.
This is the demo that makes people lean in: you ask for "a board-ready Series A deck for a climate-tech startup" or "a three-statement financial model with a five-year projection," and a minute later you have a real PowerPoint or Excel file in your hands - not a screenshot, not a markdown table pretending to be a spreadsheet, but a genuine OpenXML document you can open in Microsoft Office, edit, and send to your board.
Real files, no Office required. The agent drives OfficeCLI - billed as the first Office suite built for AI agents - which creates and edits .docx, .xlsx, and .pptx files headlessly, with a 350-plus-function Excel engine, native pivot tables, and OpenXML validation. Everything happens in a sandboxed shell, so the agent builds documents the way a developer would at a terminal, then saves the finished artifact into its space for you to download.
Skills load live from GitHub - nothing is baked in. The agent does not memorise OfficeCLI's syntax. Its know-how lives in the OfficeCLI skills repository and loads on demand: a git skills toolkit lets it list the available skills - pitch deck, financial model, data dashboard, academic paper, fillable Word form, morphing PowerPoint - and fetch the exact SKILL.md for the task, then follow it step by step. Update the repo and the agent's abilities update with it; point the two skill abilities at your own repo to ship your own document recipes.
The shell is not optional - it is the whole point. A skill is just instructions: it tells the agent which OfficeCLI commands to run to build a deck or a model. Those instructions are inert without a shell to run them in. That is why this blueprint installs the shell tools alongside the skill loader - execute commands, read and write and replace files, and import the OfficeCLI binary from a URL. Skills provide the recipe; the shell is the kitchen. Take the shell away and every skill becomes a page of commands with nothing to run them.
Why it is a killer use-case. Document generation is the task every knowledge worker actually needs and almost no chatbot can truly do. Most "make me a deck" tools hand back markdown or a flat image. This one hands back the file - formatted, formula-driven, editable, on brand. Wire the same agent into Slack, email, or a web widget and you have on-demand decks, reports, and models wherever your team already works.
Backstory
Common information about the bot's experience, skills and personality. For more information, see the Backstory documentation.
Skillset
This example uses a dedicated Skillset. Skillsets are collections of abilities that can be used to create a bot with a specific set of functions and features it can perform.
Install Shell Tools
Installs the sandboxed shell tools into the conversation - execute OfficeCLI commands and scripts, read, write, and replace files, and import the OfficeCLI binary from a URL. Required for any skill to actually build a document.List OfficeCLI Skills
List the document-building skills available in the OfficeCLI repository - returns the name, description, and path for each.Fetch OfficeCLI Skill File
Fetch a file from the OfficeCLI repository by path - use a SKILL.md path from the list ability to load a skill.
Terraform Code
This blueprint can be deployed using Terraform, enabling infrastructure-as-code management of your ChatBotKit resources. Use the code below to recreate this example in your own environment.
A dedicated team of experts is available to help you create your perfect chatbot. Reach out via or chat for more information.
