What should a reusable Claude handoff file contain before another assistant takes over?
一份可复用的Claude交接文件应包含目标、当前状态、权威文件和链接、约束、带理由的决策、失败的方法、已完成的验证、开放问题以及下一个具体步骤。区分事实与假设，并注明哪些输出可以安全覆盖至关重要。这种结构旨在防止在不同助手之间切换时重复工作和丢失上下文，同时也提出了一个问题：人可读的Markdown文件是否足够，或者最终是否需要结构化的JSON文件。
Copying a full Claude conversation preserves too much noise, while a short summary often drops the decisions that matter. A reusable handoff file could separate the goal, current state, authoritative files and links, constraints, decisions with rationale, failed approaches, validation already completed, open questions, and the next concrete step. It should also distinguish facts from assumptions and note which outputs are safe to overwrite. For people who move work from Claude to another assistant or back again, which fields consistently prevent repeated work or lost context? Does a human-readable Markdown file hold up, or do you eventually need structured JSON alongside it?