What it does SenseNova-Skills is a collection of modular “skills” for agent runtimes like OpenClaw and hermes-agent. Each skill lives in its own directory and declares triggers, capabilities, and execution flow via a SKILL.md file following the Agent Skills convention. The four main buckets: image generation and infographics, PowerPoint creation, Excel data analysis, and multi-step deep research. You can use them standalone or chain them into end-to-end workflows.

The interesting bit The project treats office work as an orchestration problem, not a chat problem. The PPT pipeline, for example, has separate creative mode (full-page PNG per slide) and standard mode (HTML per page with VLM quality control, then PPTX export). The infographic skill auto-scores prompts, picks from 87 layouts and 66 styles, and runs multi-round generation with VLM review. There’s even a “doctor” skill per category that validates your environment and fills missing .env values interactively—admitting that setup friction is the real bottleneck.

Key highlights

• Image & Visualization: text-to-image, image recognition, infographic generation with layout/style selection, image imitation from a reference, and resume image generation.
• Presentations: unified entry point that parses PDF/DOCX/MD/TXT inputs, then dispatches to creative or standard PPT generation pipelines.
• Data Analysis: end-to-end Excel workflow with multi-sheet read, automatic Parquet conversion for ≥10k rows, cleaning, filtering, and export; plus image-to-data extraction via OCR and chart understanding.
• Deep Research: resumable execution from planning through per-dimension evidence gathering to final report.md synthesis.
• Hosted option: all skills are bundled into Raccoon, a hosted product with a free trial—useful if you don’t want to provision API keys and runtimes yourself.

Caveats

• The README is upfront that you may need to manually restart your agent service after skill installation.
• Designed primarily for the SenseNova model family and platform API; mileage with other LLM backends is unclear.
• The “15M+ individual users” claim refers to the hosted Raccoon product, not this open-source repo itself.

Verdict Worth a look if you’re building agent-based office automation and want pre-built, composable skills rather than wiring APIs yourself. Skip it if you’re not already running an Agent Skills-compatible runtime or if your use case is outside document-heavy knowledge work.