OpenSenseNova/SenseNova-Skills

Adopt SenseNova-Skills only if you already run a compatible agent runtime and are willing to route work through SenseNova models; it is an extension pack, not a standalone assistant.

The problem

A generic agent can plan and talk, but ask it to produce a deck, analyze a spreadsheet, or run a staged research report and you end up hand-building prompts and API calls for each task. SenseNova-Skills supplies those task recipes as named skills, so the agent invokes a defined workflow instead of improvising. The change is at the invocation step: the runtime loads a skill, the skill calls the documented SenseNova APIs, and the user gets a finished artifact rather than a chain of ad-hoc prompts.

What's different

The repository organizes capabilities by office outcome: image and visualization, presentations, data analysis, deep research, search, and scroll-driven HTML storytelling. Each area has its own guide covering prerequisites, Quick Start, API configuration and invocation samples, and one skill, sn-motion-html, ships its SKILL.md in the repo with project setup and browser QA notes. That per-skill guide structure is the practical benefit: you can read one document to judge whether a single workflow fits before wiring the rest.

The trade-off is coupling. The README recommends pairing the skills with OpenClaw or hermes-agent and with the SenseNova Platform API, whose base URLs differ for international and mainland China access. The skills are not described as a general-purpose plugin format that drops into any assistant; they assume a skills-compatible runtime and SenseNova credentials. If you use a different model provider, the documented path does not cover you.

Who should try it

Good fit: a team already using OpenClaw or hermes-agent that needs repeatable office output, such as a research brief with staged deep-research steps or a presentation generated from source material, and that is comfortable using SenseNova models. The staged deep-research documentation is the strongest reason to look here, because multi-step research is where ad-hoc prompting usually drifts.

Poor fit: someone who wants a ready-made assistant with no runtime to set up, or who has standardized on another model vendor. Those users should look at the hosted option in the README instead of the repo. A developer building a single custom workflow may also find a full skills runtime heavier than writing the API calls directly.

Requirements and limitations

You need a skills-compatible agent first; the skills are not useful on their own. The recommended runtimes are named, and the README does not promise compatibility with arbitrary agents. You also need SenseNova Platform API access and keys, with separate documentation for international and mainland China endpoints.

Operationally, some skills depend on additional per-platform API keys; the search skills, for example, are documented alongside deep research with their own key requirements. The FAQ covers 400/401 errors, rate limits, PPT timeouts and infographic quality, which signals that these are real setup and runtime concerns rather than edge cases. Expect to maintain model names and endpoint configuration as the platform changes.

HotRepo's take

HotRepo's recorded sample is thin: two daily-board observations, 2026-10-24 at 5,898 stars (+26) and 2026-10-25 at 5,900 stars (+19). That is not enough comparable history to judge whether interest is rising or fading, so persistence cannot yet be assessed. The repo was 193 days old at the sample date, so treat the star count as a snapshot rather than a maturity signal.

Before investing, verify one thing in the docs: whether your chosen runtime is supported beyond the two recommended ones, since that determines whether this pack fits your existing setup.