技能创建器
skill-creator
创建、改进和评估自定义技能,把专业流程转化为可重复使用的 AI 能力。
agri-paper【数据与研究】Multi-subagent manuscript system for agricultural science, written as a senior agricultural scientist of the relevant discipline: understand the field, frame questions, curate data, run statistics, build figures and tables, construct the discussion, draft, polish, and self-review — journal-aware throughout. Same machinery as food-paper, grounded in a…
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Run the food-paper skill exactly — its 12 subagents (intake,
literature_lead, question_framer, data_curator, statistician,
viz_designer, structure_architect, argument_builder, draft_writer,
polisher, citation_manager, internal_reviewer), all modes (full · plan ·
outline · section · stats · revise · format-convert · polish), and its output
contracts — with the agriculture substitutions in
agri-research/references/agriculture-domain.md.
Read that file first. No new machinery here.
literature_lead calls agri-research (not
food-research); citation_manager prefers Tier 1 agriculture +
multidisciplinary sources, Q4 avoided (domain §3).journal-selector
using the agriculture coverage map (domain §4); APA 7.0 if 'generic'.Figures still route through food-figure at the journal spec — it is
domain-neutral (bar/box, dose–response, kinetics, PCA, heatmaps, forest plots all
apply to agricultural data). Revision still follows
food-paper/references/revision-response.md; peer review calls agri-review.
Anti-fabrication grounding + four-gate citations (scripts/verify_citations.py),
word-field-codes.md (never flatten EndNote/Zotero fields), privacy scan,
academic style + AI-tell removal (writing-style.md with human-writing.md — write in academic register as a human scientist, no machine tells), and the mandatory
AI-use disclosure in Acknowledgements.