dsh-academy
# Academic mode for DeepSeek Harness English | [中文](README_CN.md) [](https://www.npmjs.com/package/dsh-academy)  One small plugin that changes how the agent *behaves* on scholarly work: a research persona, the anti-fabrication rules that matter most in academic writing, and bilingual conventions. It registers **no tools and no skills** — capabilities belong to other plugins. This is the behaviour layer that makes them read like a research assistant. ```sh dsh plugin --profile web add dsh-academy dsh web ``` ## What it adds to the system prompt **Research persona** — the reader is a researcher who needs something that survives peer review, not a paragraph that sounds right. **Academic-integrity rules** — never invent papers, authors, years, journals, DOIs or citation counts; never invent data; separate what came from the literature from what is your inference; make every citation checkable; leave judgement calls to the researcher. **Language conventions** — answer in the user's language, keep titles, journal names and terminology in the original, follow Chinese academic phrasing when writing Chinese. ## Configuration ```yaml - id: academy config: persona: true integrity: true language: true field: "" # e.g. 分子生物学 — named in the persona personaOrder: 0 # the deployment-persona slot rulesOrder: 120 ``` ## Works well with - [`dsh-ai4scholar`](https://github.com/literaf/ai4scholar-plugin-dsh) — 38 literature tools (Semantic Scholar, PubMed, Google Scholar, arXiv, bioRxiv/medRxiv, DOI full text, auto-cite, figures). The integrity rules are only enforceable when the agent can actually look papers up. - [`dsh-research`](https://github.com/literaf/dsh-research) — the research skill pack (peer review, introductions, formatting, reference audit). - [`dsh-research-hub`](https://github.com/literaf/dsh-research-hub) — the index of research plugins for dsh. ## License MIT