A 29-Skill Office Suite for OpenClaw and Hermes Agents: SenseNova-Skills

Most agent skills do one trick. SenseNova-Skills — from the SenseNova model family, built to plug into agent runtimes like OpenClaw and hermes-agent — ships 29 skills across eight areas: image generation and visualization, slide-deck generation, Excel data analysis, deep research, search, HTML experiences, team collaboration, and proactive project tracking. Each lives in its own directory with a SKILL.md following the Agent Skills convention, and the skills are designed to compose: data analysis feeds deep research, which feeds the PPT pipeline.

Why This Skill Matters

The PPT chain is the clearest example of the composition design. sn-ppt-entry collects role, audience, scene, page count, and mode (standard, dazzle, or creative), parses pdf/docx/md/txt inputs, and emits a task pack. sn-ppt-story — which the repo marks as mandatory and not to be skipped or hand-written — turns the materials into a single editable outline. sn-ppt-standard then runs style spec, asset plan, per-slot images with VLM quality checks, per-page HTML, and a review pass before exporting PPTX. The dazzle mode renders a single-file 1280×720 dynamic HTML deck with motion and keyboard navigation instead.

The data-analysis side handles real office shapes: sn-da-excel-workflow reads multi-sheet files, auto-detects large files (10k+ rows trigger Parquet conversion), cleans, filters, aggregates across sheets, and exports Excel or CSV. sn-deep-research orchestrates parallel research work packages into an auditable workflow ending in report.md, with academic and social search skills (ArXiv, Semantic Scholar, PubMed, Bilibili, Zhihu, and more) feeding it evidence.

Installation

The recommended path is to let your agent do it — hand it this request; the README says you may need to manually restart the agent service afterwards before the new skills get picked up:

Please install SenseNova-Skills from https://github.com/OpenSenseNova/SenseNova-Skills into your skills directory.

Manual install for OpenClaw (use ~/.hermes/skills/ for Hermes):

git clone https://github.com/OpenSenseNova/SenseNova-Skills.git --depth=1
mkdir -p ~/.openclaw/skills
cp -r SenseNova-Skills/skills/* ~/.openclaw/skills/

Skills need a model behind them: the repo recommends pairing with the SenseNova Platform API, and warns that the docs page, API key, base URL, and model name must all come from the same region — international (https://token.sensenova.ai/v1) or mainland China (https://token.sensenova.cn/v1).

Real Workflow: From Raw Spreadsheet to Board-Ready Deck

1. Install the suite and confirm the skills load in a fresh session.
2. Hand the agent a spreadsheet and a goal:

把这份各月销售明细 Excel 分析一遍：先清洗数据，按地区和品类做交叉汇总，
找出环比变化最大的三个品类，输出 Excel 报表。

3. sn-da-excel-workflow reads every sheet, cleans and joins the data, and returns the aggregates plus the export — a 10k-row file switches to Parquet-backed processing automatically.
4. Feed the findings into a deck:

基于刚才的分析结论，用 sn-ppt-entry 生成一份 12 页的 standard 模式汇报 PPT，
受众是管理层，风格正式一些。

5. The pipeline runs entry → story outline → standard mode, and you review the editable outline.md before the pages render — that outline is the checkpoint where you redirect the narrative.
6. The repo's own end-to-end example (memory-price analysis) shows the full arc: a raw quote CSV becomes trend analysis, then a researched report, then a 16-page PPTX.

Tips

• Want a single visual instead of a full deck? sn-infographic generates publication-ready infographics with auto prompt-quality scoring, 87 layouts and 66 styles, and VLM review with quality ranking.
• Setup problems have a map: sn-image-doctor and sn-ppt-doctor check environment, dependencies, and render/export tooling, and the FAQ covers 400/401 errors, rate limits, and PPT timeouts.
• Prefer zero setup? The README notes the same model and skill suite ships in the Raccoon office product as a free trial, with no environment to provision.
• If search skills return nothing, check platform auth — some Chinese social platforms need cookie authentication, per the search skill table.

When Not to Use This

The suite assumes the SenseNova API as its recommended model and OpenClaw or Hermes as runtimes — on other setups, per-category Python dependencies and API configuration are on you, and the office-product route (Raccoon) may be simpler than self-hosting. And if your agent already has a battle-tested slide or research skill, this suite's value is the composition — using one piece in isolation forgoes most of it.

See the leaderboard for more skills.