literaf/dsh-academy
dsh-academy adds an academic behaviour layer to DeepSeek Harness: it registers no tools or skills, instead contributing a research persona, anti-fabrication integrity rules (never invent papers, authors, years, journals, DOIs or data; separate literature from inference) and bilingual academic language conventions to the system prompt.
dsh-academy adds an academic mode to DeepSeek Harness by changing how the agent behaves on scholarly work rather than what it can do. It contributes three system-prompt layers: a research persona (the reader is a researcher who needs something that survives peer review), 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 inference; make every citation checkable; leave judgement calls to the researcher) and bilingual language conventions (answer in the user's language, keep titles, journal names and terminology in the original, follow Chinese academic phrasing when writing Chinese).
Install
dsh plugin --profile web add dsh-academyCopied verbatim from the README's install block (the following line is dsh web).
Compatibility
A DeepSeek Harness behaviour-layer plugin: it registers no tools and no skills - capabilities belong to other plugins. Its config toggles persona, integrity and language sections independently and takes an optional field and ordering slots. The README pairs it with the sibling dsh-ai4scholar (38 literature tools).
Details
- Repo: literaf/dsh-academy
- Category: Learning & Research
- Stars: 0
- Version: GitHub repo literaf/dsh-academy (0★, MIT); npm package dsh-academy at 0.1.0 (registry check 2026-10-14; the registry repository field resolves back to this repo).
Alternatives
1692775560/dsh-Mimir-Academic-research · Astervolans/dsh-literature-search · beibingyangliuying/dsh-autoresearch