Craft practices rely on evolving archives of skill and knowledge developed through generations of craftspeople experimenting with designs, materials, and techniques. Better documentation of these practices enables the sharing of knowledge and expertise between sites and generations. However, most documentation focuses on the linear steps leading to final artifacts, neglecting the distinct tacit knowledge, improvisational actions, and situated adaptations needed to meet the unique demands of each craft project. This omission limits knowledge sharing and reduces craft to a mechanical endeavor, rather than a sophisticated and contextual way of seeing, thinking, and doing. Drawing on expert interviews and literature from HCI, CSCW and the social sciences, we develop an elementary grammar to document improvisational actions of real-world craft practices. We demonstrate the utility of this grammar with a MLLM-powered interface called CraftLink that can be used to analyze expert videos and generate documentation to share material and contextual variations of practices with other knowledgeable but non-master craftspeople. Our user study with expert crocheters (N=7) evaluates our grammar's effectiveness in capturing and sharing expert knowledge with other craftspeople, offering new pathways for computational systems to support collaborative archives of knowledge and practice across time, space, and skill levels. We conclude by showing how our grammar address four key tensions of the craft learning environment: personal and shareable documentation, fragmented and discoverable expertise, linear and iterative practices, and data privacy and ownership.

翻译：手工艺实践依赖于历代工匠通过设计、材料与技术实验所形成的不断演进的技能与知识档案。对这些实践进行更完善的记录，能够促进不同地域与代际间的知识与专长共享。然而，现有记录大多聚焦于最终成品的线性制作步骤，忽视了每个手工艺项目为满足其独特需求所必需的隐性知识、即兴操作与情境化调整。这种缺失限制了知识共享，并将手工艺简化为机械性劳动，而非一种精妙且情境化的观察、思考与行动方式。基于专家访谈以及来自人机交互、计算机支持的协同工作与社会科学的文献，我们提出一种基础语法，用于记录真实手工艺实践中的即兴操作。我们通过一个名为 CraftLink 的 MLLM 驱动界面展示了该语法的实用性：该界面可分析专家视频并生成记录文档，从而与其他具备相关知识但非大师级的工匠共享实践中的材料与情境变体。我们通过对专业钩针编织者（N=7）的用户研究，评估了该语法在获取并向其他工匠分享专家知识方面的有效性，为计算系统支持跨越时间、空间与技能水平的协作式知识与实践档案提供了新路径。最后，我们展示了该语法如何应对手工艺学习环境中的四组核心矛盾：个人化与可共享的记录、碎片化与可发现的专长、线性与迭代的实践，以及数据隐私与所有权。