Plug-in office skills for AI agents that actually do the work

What it does This repository is a collection of drop-in agent skills built around the SenseNova model family. Each skill lives in its own directory and declares its triggers and execution flow via a SKILL.md file under the Agent Skills convention. The modules cover four office territories: image generation and visualization, PowerPoint creation, Excel data analysis, and deep research workflows.

The interesting bit The suite is deliberately tiered. A Tier 0 base layer handles low-level image and text operations, while Tier 1 skills orchestrate complex pipelines—like infographic generation that auto-scores prompts and picks from 87 layouts, or Excel workflows that automatically pivot to Parquet when files exceed 10,000 rows. Everything is designed to be called by an upper-layer agent rather than used as a standalone app.

Key highlights

• PPT generation offers two modes: a “creative” path that emits full-page 16:9 PNGs per slide, and a “standard” path that runs through outline → asset plan → HTML → review → PPTX export.
• Data analysis skills include multi-sheet Excel orchestration, image-first table OCR and chart understanding, and streaming large-file handling via chunked Parquet conversion.
• Deep research provides a resumable end-to-end pipeline: planning, per-dimension evidence gathering, synthesis, and final report.md output.
• Infographic generation leverages the SenseNova U1 model with automated prompt-quality scoring, VLM review, and a library of 87 layouts and 66 styles.
• The same skill suite ships inside Raccoon, SenseNova’s hosted office assistant, which claims 15 million+ individual users.

Caveats

• The skills are tightly coupled to the SenseNova ecosystem; the README recommends pairing them with the SenseNova Platform API and runtimes like OpenClaw or hermes-agent.
• Several capabilities—especially image generation and deep research—require specific API keys and external dependencies whose exact scope is only detailed in separate per-category docs.

Verdict Worth exploring if you are already running an Agent Skills-compatible runtime and want pre-built, opinionated office automations. If you are not invested in the SenseNova model stack or the Agent Skills convention, this is mostly a reference architecture for how to structure agent plugins.