SenseNova-Skills: a suite of end-to-end office capabilities for AI agents including infographic generation and deep research

What it solves

SenseNova-Skills provides a suite of end-to-end office capabilities for AI agents, enabling them to perform complex professional tasks such as generating publication-ready infographics, creating structured slide decks (PPT), analyzing large Excel datasets, and conducting deep, evidence-based research.

How it works

The project follows the Agent Skills convention, where each skill is a standalone directory containing a SKILL.md file that defines its triggers, capabilities, and execution flow. These skills are designed to plug into agent runtimes like OpenClaw or hermes-agent and are optimized for use with the SenseNova model family. The system operates across several specialized tiers:

• Image & Visualization: Uses models like SenseNova U1 to generate infographics with automated layout selection and VLM (Vision Language Model) quality review.
• Presentations: Orchestrates a pipeline from outline and asset planning to HTML rendering and final PPTX export.
• Data Analysis: Handles multi-sheet Excel reading, large-file processing via Parquet conversion, and image-to-data extraction.
• Deep Research: Implements a structured loop of planning, per-dimension evidence gathering via web search, synthesis, and final report writing.

Who it’s for

It is designed for developers building AI agents and enterprise users who need to automate high-value office workflows, such as market research, financial analysis, and professional presentation design.

Highlights

• End-to-End Office Suite: Covers the full loop from raw data analysis to research and final presentation.
• Infographic Generation: Capable of producing dense, structured visual infographics using 87 layouts and 66 styles.
• High-Performance Data Analysis: Supports streaming reads for Excel files with over 10,000 rows using memory optimization.
• Structured Research: Replaces simple search with a plan-then-execute loop that cross-validates evidence across multiple sources.
• Multi-Runtime Support: Compatible with OpenClaw and hermes-agent.