literaf/dsh-research

README 摘要

dsh-research AI4Scholar Research Pack for DeepSeek Harness English 中文 The orchestration layer that turns DeepSeek Harness into a research workbench. It searches nothing itself: it ships the research skills, the literature-first workflow guidance, and the workspace conventions that make capability plugins ( dsh-ai4scholar ) and the agent behave like one research assistant. What it installs Four research skills (authored by AI4Scholar, embedded in the package — nothing to download): Skill When the agent loads it ai4scholar-paper-review Pre-submission review from five reviewer perspectives, with a fix-priority list ai4scholar-introduction-writing Introductions: background → gap → contribution → signposting → polish ai4scholar-academic-formatting Reference formatting, three-line tables, abstract structure, re-submission reformatting ai4scholar-reference-audit Reference audit: cross-citation, format consistency, DOI checks, duplicates, self-citation One workflow section in the system prompt: search before asserting, which task loads which skill, and a workspace layout — papers in papers/ , citations in references.bib , reading notes in notes/ — so the next session can pick up where the …

分类