dsh-memory-evolveA72

The package.json declares no dsh.bundle, so even once installed into a profile it is never assembled. It may be an index list, documentation, a standalone app, or still in development — check the repo README for how to use it.

An AI conversation is normally one-shot: switch projects, come back a few days later, open a new session, and it has forgotten who you are. This plugin gives DSH a long-term working memory that spans sessions. Memory is split into five tracks: a user profile (your preferences, company and communication habits, visible every turn), global facts (environment, tools, conventions and other durable knowledge), project key memory (the current project's conventions, decisions, architecture and pitfalls, injected into context automatically, with important conclusions optionally scoped to a single git branch), and project and daily logs (progress recorded automatically each turn, fully traceable). Writes require your confirmation before they take effect — the AI never slips things into memory on its own — and once memory grows you can have entries archived, which removes them from context injection while keeping them one command away from returning. Your reactions to results — a "nice" or a "why still not fixed" — get recorded as `【反馈】` lines in the daily and project logs, so after a while you can ask the AI to analyze which kinds of task it handles well or badly. Installing adds a row of capability tabs — memory, skills, to-dos, infinite canvas, COI scheduling, session broadcast, prompts, memory sync, model settings, bookmarks, session review and more. Many are off by default so they don't crowd the AI's tool list; turn on what you need in settings. Memory sync shares project memory across machines, storing it by default on a dedicated branch of your own code repository so it never pollutes the code; if your code is public but your memory should stay private, point it at a separate shared memory repository instead. The four to-do tracks split life, work, project and daily, with the project track isolated per working directory.

• ▸Five memory tracks — user profile, global facts, project key memory, project log, daily log — with important conclusions scopable to a single git branch
• ▸Writes take effect only after you confirm; entries can be archived to stop context injection and restored at any time
• ▸Memory sync shares project memory across machines on a dedicated branch of your code repo, or a separate shared memory repo when code is public
• ▸Four to-do tracks (life / work / project / daily) isolated per working directory, plus COI scheduling, session broadcast and search, and skill self-evolution

五轨记忆+待办+技能+会话编排的重型记忆插件，场景化文档写得极好，9 天 288 commit

Drop it into your README — it links back to this page.

Badge (inline)Copy Markdown

[![dshfind](https://dshfind.com/api/badge/csyangwen/dsh-memory-evolve)](https://dshfind.com/en/plugins/csyangwen/dsh-memory-evolve?ref=badge)

Card (440×200)Copy Markdown

[![dshfind](https://dshfind.com/api/card/csyangwen/dsh-memory-evolve)](https://dshfind.com/en/plugins/csyangwen/dsh-memory-evolve?ref=badge)