AI agents that drive Photoshop, Office, and Zotero by COM

What it does This repository bundles several Python CLI tools that expose Windows COM automation interfaces for desktop applications—WPS/MS Office, Adobe Photoshop, Adobe Illustrator, and Zotero—so an AI agent can create documents, edit images, manage literature, and generate presentation decks through command-line calls instead of manual clicking.

The interesting bit Instead of building new APIs, it treats decades-old COM layers as the bridge between modern AI agents and stubbornly GUI-centric software, effectively turning Word and Photoshop into headless(ish) backends. The Zotero module goes further by codifying academic workflows—literature search, IMRAD writing, meta-analysis, and peer review—into 27 reusable “skills.”

Key highlights

• 47 CLI commands for document, spreadsheet, and presentation manipulation in WPS or Microsoft Office
• 27 academic skills for Zotero-backed research, from systematic reviews to journal formatting pipelines
• COM bridges for Adobe Photoshop and Illustrator enabling layer, shape, text, and export operations via CLI
• JSON-driven PPT generation with design-system presets, layout routing, and multi-format export
• Each sub-project installs separately; requires Windows 10/11 and locally installed host applications

Caveats

• Windows-only; every harness requires its target application to be installed and licensed locally
• README structure is disjointed (sections jump from 3 to 5 to 4), suggesting a still-evolving monorepo layout
• COM interop behavior will vary with each host application version, and the docs do not address error handling or recovery

Verdict Windows-bound researchers and automation builders who need AI agents to physically manipulate documents, images, and citation libraries will save months of workaround scripting. If your stack is cloud-native or cross-platform, this is a hard pass.