name: "consolidate-memory"
description: "Reflective pass over your memory files — merge duplicates, fix stale facts, prune the index."
Memory Consolidation / 记忆整合
You're doing a reflective pass over what you've learned about this user and their work. The goal: a future session should be able to orient quickly — who they work with, what they're focused on, how they like things done — without re-asking.
你正在对已了解的该用户及其工作进行一次回顾性梳理。目标是:未来的会话应能快速进入状态——了解用户与谁共事、正在专注什么、偏好怎样的做事方式——而无需重新询问。
Your system prompt's auto-memory section defines the directory, file format, and memory types. Follow it.
系统提示词中的自动记忆(auto-memory)部分定义了目录、文件格式和记忆类型,请遵循该部分的规定。
Phase 1 — Take stock / 第一阶段——盘点
- List the memory directory and read the index (
MEMORY.md)- 列出记忆目录并读取索引(
MEMORY.md)
- 列出记忆目录并读取索引(
- Skim each topic file. Note which ones overlap, which look stale, which are thin.
- 粗读每个主题文件,记下哪些内容重叠、哪些已过时、哪些过于单薄。
Phase 2 — Consolidate / 第二阶段——整合
Separate the durable from the dated. Preferences, working style, key relationships, and recurring workflows are durable — keep and sharpen them. Specific projects, deadlines, and one-off tasks are dated — if the date has passed or the work is done, retire the file or fold the lasting takeaway (e.g. "user prefers X format for launch docs") into a durable one.
**区分持久信息与时效信息。**偏好、工作风格、关键人际关系和周期性工作流属于持久信息——保留并提炼它们。具体项目、截止日期和一次性任务属于时效信息——如果日期已过或工作已完成,就废弃该文件,或把有长期价值的要点(例如"用户偏好发布文档采用 X 格式")并入持久类文件。
Merge overlaps. If two files describe the same person, project, or preference, combine into one and keep the richer file's path.
**合并重叠内容。**如果两个文件描述的是同一个人、项目或偏好,则合并为一个,并保留内容更丰富文件的路径。
Fix time references. Convert "next week", "this quarter", "by Friday" to absolute dates so they stay readable later.
**修正时间表述。**把"下周"、"本季度"、"周五前"转换为绝对日期,以便日后阅读时仍然可理解。
Drop what's easy to re-find. If a memory just restates something you could pull from the user's calendar, docs, or connected tools on demand, cut it. Keep what's hard to re-derive: stated preferences, context behind a decision, who to go to for what.
**舍弃易于重新获取的内容。**如果某条记忆只是复述了随时可以从用户日历、文档或已连接工具中查到的信息,就删掉它。保留难以重新推导的内容:明确表达的偏好、决策背后的背景、什么事该找什么人。
Phase 3 — Tidy the index / 第三阶段——整理索引
Update MEMORY.md so it stays under 200 lines and ~25KB. One line per entry, under ~150 chars: - [Title](file.md) — one-line hook.
更新 MEMORY.md,使其保持在 200 行以内、约 25KB 以内。每条占一行、不超过约 150 个字符:- [Title](file.md) — one-line hook。
- Remove pointers to retired memories
- 移除指向已废弃记忆的条目
- Shorten any line carrying detail that belongs in the topic file
- 缩短任何承载了本应属于主题文件细节的行
- Add anything newly important
- 补充新近变得重要的内容
Finish with a short summary: how many files you touched and what changed.
最后给出一段简短总结:你改动了多少个文件以及发生了什么变化。