dsh-mnemon 0.5.33 → 0.5.34

package/README.md

@@ -12,8 +12,8 @@
   <a href="https://dshfind.com/en/plugins/omdsh-dev/dsh-mnemon?ref=badge"><img alt="dshfind downloads" src="https://dshfind.com/api/badge/omdsh-dev/dsh-mnemon?metric=downloads&amp;lang=en" /></a>
 </p>
 
-<p align="center"><strong>Composable, view-based memory for DeepSeek Harness.</strong></p>
-<p align="center">Pluggable sources and strategies, with layered memory out of the box.</p>
+<p align="center"><strong>Composable, view-based context for DeepSeek Harness.</strong></p>
+<p align="center">Pluggable sources and strategies, with layered memory and self-evolving skills out of the box.</p>
 
 <p align="center">
   <a href="https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/en/guides/ui-guide.md#in-a-conversation">
@@ -30,11 +30,12 @@
 
 ## Why dsh-mnemon
 
-An Agent that starts every session from scratch keeps asking what it was already told. Putting everything into one memory store is no better: it either floods each turn or misses what matters. dsh-mnemon gives DeepSeek Harness memory that is layered, visible and composable.
+An Agent that starts every session from scratch keeps asking what it was already told. Putting everything into one memory store is no better: it either floods each turn or misses what matters. dsh-mnemon gives DeepSeek Harness memory that is layered, visible and composable, and skills that improve with use.
 
 - **The right memory for each turn.** Preferences and working facts stay in context. Project documents and long-term evidence are searched only when a question needs them.
 - **You can see what a turn used.** Under each reply, the turn memory bar lists the documents and memories the turn read and wrote, and one click opens each in the Memory System. The Memory System shows everything that is stored and lets you edit it.
 - **Save to memory in one click.** When a conversation turns up a fact worth keeping, the brain mark under the reply hands it to a task Agent, which removes duplicates, distils it and writes it to the right Memory Space, or saves it where you choose: working memory, your profile or a Memory Space. The receipt says where it went.
+- **Skills that improve with use.** Skills keeps every DSH skill in one place. It can review your real conversations, draft a better `SKILL.md` and backtest it before you adopt it, and it installs more skills from skills.sh, GitHub and sites that publish Agent Skills.
 - **Compose it on the Plugins page.** Choose a main strategy and optional enhancements, and switch them without moving any data. Every component has its own page for its settings.
 - **Keep data where you want it.** Local by default with Mnemon Native, or one of eight third-party Providers. Storage can be global, per workspace, centralized or in a directory you choose, with ZIP backup.
 - **Extend it.** Sources and Strategies are ordinary DSH plugins built on public SDKs. An installed component gets the same pages and switches as the shipped ones.
@@ -73,6 +74,22 @@ Every step is explained in the [UI guide](https://github.com/omdsh-dev/dsh-mnemo
 
 The default **Layered strategy** keeps runtime memory resident and reads the other two on demand. The **General strategy** offers every available source within one budget and lets the model decide how to use each. A **memory space** is one named, Provider-backed scope for long-term evidence; the Chinese product term is 记忆空间.
 
+## Skills that improve with use
+
+**Skills** is an experimental component for the `SKILL.md` skills DSH loads from your project and your home folder. Its page sits in the Memory System, or in the sidebar on its own.
+
+| A turn's review suggests an improvement | You adopt it after a backtest |
+|---|---|
+| [![The Turn skills line under a reply: release-notes, suggested 1](https://raw.githubusercontent.com/omdsh-dev/dsh-mnemon/main/docs/assets/skills-v0.5.34/en/chat-skills.jpg)](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/en/guides/skills.md#in-a-conversation) | [![A draft of release-notes with its backtest, Current 40 → Draft 93 +53, and the diff](https://raw.githubusercontent.com/omdsh-dev/dsh-mnemon/main/docs/assets/skills-v0.5.34/en/skills-draft.jpg)](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/en/guides/skills.md#evolution) |
+
+- **Manage.** Switch a skill off for the model or the `/` menu, use it in the conversation in one click, edit it, and go back to any earlier version.
+- **Evolve.** With self-evolution on, finished conversations are reviewed in the background for corrections, "from now on" instructions, failed tools and reusable procedures. A suggestion cites the steps it rests on; a task Agent drafts the new `SKILL.md`, which is backtested against the current version before you adopt it. Nothing changes without your **Adopt**.
+- **Test.** Each skill has its own cases, from reviews, from recorded conversations or written by hand, and a backtest scores any version on them.
+- **Install.** The market searches skills.sh's open index and reads any GitHub repository, any site publishing an Agent Skills discovery index, and curated catalogs; every install is previewed, with security reviews where they exist.
+- **In the conversation.** A line under each turn names the skills it used and what its review suggested, each one click from its page.
+
+Self-evolution stays off until you turn it on under **Plugins → dsh-mnemon → Skills**. The [Skills guide](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/en/guides/skills.md) walks through every screen.
+
 ## Quick start
 
 New to DSH? [Install and start](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/en/guides/installation.md) walks from an empty machine to your first memory, with a screenshot for every step. The short version, with [Node.js](https://nodejs.org/) 22.19 or later:
@@ -113,6 +130,7 @@ The Starter pins a tested combination of independently versioned packages. One m
 | [dsh-mnemon-source-runtime](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-source-runtime/README.md) | Runtime memory: USER.md, MEMORY.md, revisions and local storage | On |
 | [dsh-mnemon-source-documents](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-source-documents/README.md) | Project Documents: Markdown, search, revisions and archiving | On |
 | [dsh-mnemon-source-memory-spaces](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-source-memory-spaces/README.md) | Memory Spaces: durable evidence and their Providers | On |
+| [dsh-mnemon-source-skills](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-source-skills/README.md) | Skills (experimental): DSH skills with versions, self-evolution from real conversations, tests and backtests, and a market of skills.sh, GitHub and Agent Skills sites | On (self-evolution off) |
 | [dsh-mnemon-strategy-default-three-tier](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-strategy-default-three-tier/README.md) | Layered strategy: resident runtime memory, the rest on demand | Selected |
 | [dsh-mnemon-strategy-general](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-strategy-general/README.md) | General strategy: every available source in one budget | Off |
 | [dsh-mnemon-strategy-auto-capture](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-strategy-auto-capture/README.md) | Active capture: in-turn guidance to keep useful facts | Off |
@@ -131,6 +149,7 @@ Start with the [plugin author guide](https://github.com/omdsh-dev/dsh-mnemon/blo
 
 - Runtime memory and Project Documents are local files; Mnemon Native is local. Third-party Providers use their own services and scopes.
 - Turning a component off does not erase its memory, and changing the storage location does not move data. Use ZIP backup to carry it.
+- Skills keeps its evolution data (suggestions, test cases and the conversations reviews read, with secrets redacted) in the local storage root, outside backups. An Agent follows a skill from the market as written, so preview it before installing.
 - Git sync publishes the same Mnemon Pack payload as readable files to a repository you own, so another machine can pull it, with **Sync now**, **Check remote** before anything is written, or **Automatic sync**. When two machines wrote one subject differently, both wordings stay until you accept and apply a **Memory reconciliation** plan; see [Git repository sync](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/en/guides/operations.md#git-repository-sync).
 - Saved Provider credentials stay on the Host and are never exported. Backups still contain private memory; protect them.
 - Sources and Strategies are trusted in-process JavaScript, not sandboxed code. Remembered history never outranks current instructions.

package/README.zh-CN.md

@@ -13,8 +13,8 @@
 </p>
 
 
-<p align="center"><strong>面向 DeepSeek Harness 的可组合视图记忆。</strong></p>
-<p align="center">记忆来源与策略可插拔，开箱即用提供分层记忆。</p>
+<p align="center"><strong>面向 DeepSeek Harness 的可组合视图上下文。</strong></p>
+<p align="center">来源与策略可插拔，开箱即用提供分层记忆与自进化技能。</p>
 
 <p align="center">
   <a href="https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/zh-CN/guides/ui-guide.md#在对话中">
@@ -31,11 +31,12 @@
 
 ## 为什么需要 dsh-mnemon
 
-每次会话都从零开始的 Agent，会反复询问你早已说过的事情；把所有内容塞进同一个记忆库也不行，要么每轮都被淹没，要么漏掉真正要紧的信息。dsh-mnemon 为 DeepSeek Harness 提供分层、可见、可组合的记忆。
+每次会话都从零开始的 Agent，会反复询问你早已说过的事情；把所有内容塞进同一个记忆库也不行，要么每轮都被淹没，要么漏掉真正要紧的信息。dsh-mnemon 为 DeepSeek Harness 提供分层、可见、可组合的记忆，以及越用越好的技能。
 
 - **每一轮都拿到合适的记忆。** 偏好和工作中的事实常驻上下文；项目档案与长期证据只在问题需要时才检索。
 - **看得见这一轮用了什么。** 每条回复下方的回合记忆栏列出这一轮读到和写入的档案与记忆，点一下就在记忆系统中打开它；记忆系统展示全部已保存的内容，并可直接编辑。
 - **一键存入记忆。** 对话里出现值得保留的事实，点击回复下的脑形图标，由任务 Agent 去重、提炼并写入合适的记忆空间，也可以直接存到你选择的位置：工作记忆、用户画像或某个记忆空间。回执写明存到了哪里。
+- **越用越好的技能。** 技能管理把 DSH 的技能集中到一处：它能复盘你的真实对话，写出更好的 `SKILL.md`，先回测再交给你采纳；还能从 skills.sh、GitHub 和发布 Agent Skills 的网站安装更多技能。
 - **在插件页组合。** 选择一个主策略和若干可选增强，切换时无需迁移任何数据；每个组件都有自己的设置页。
 - **数据放在你想放的地方。** 默认由 Mnemon Native 在本地保存，也可以接入八种第三方 Provider；存储范围可选全局、按工作区、集中存储或你指定的目录，并支持 ZIP 备份。
 - **可以扩展。** Source 与 Strategy 都是基于公开 SDK 的普通 DSH 插件；安装的组件与随附组件拥有同样的页面和开关。
@@ -74,6 +75,22 @@
 
 默认的**分层策略**让运行时记忆常驻，另外两类按需读取；**通用策略**在同一份预算内提供全部可用来源，由模型决定如何使用。**记忆空间**是由 Provider 承载、可以独立命名和激活的长期证据范围，英文界面称 memory space。
 
+## 越用越好的技能
+
+**技能管理**是一个实验性组件，用来管理 DSH 从项目和用户目录加载的 `SKILL.md` 技能。它的页面在记忆系统里，也可以作为独立的侧边栏入口。
+
+| 回合复盘提出改进 | 回测之后由你采纳 |
+|---|---|
+| [![回复下方的本回合技能一行：release-notes · 建议 1](https://raw.githubusercontent.com/omdsh-dev/dsh-mnemon/main/docs/assets/skills-v0.5.34/zh-CN/chat-skills.jpg)](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/zh-CN/guides/skills.md#对话中) | [![release-notes 的草案：回测“当前 40 → 草案 93 +53”与差异](https://raw.githubusercontent.com/omdsh-dev/dsh-mnemon/main/docs/assets/skills-v0.5.34/zh-CN/skills-draft.jpg)](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/zh-CN/guides/skills.md#进化) |
+
+- **管理。** 让技能不被模型加载或不出现在 `/` 菜单里，一键在对话中使用，直接编辑，并能回到任意历史版本。
+- **进化。** 开启自进化后，对话结束会在后台复盘，寻找多次纠正、“以后都这样做”的指示、工具失败和可复用的流程。每条建议都标出所依据的步骤；任务 Agent 写出新的 `SKILL.md`，先与当前版本回测对比，再由你采纳。没有你的**采纳**，什么都不会改。
+- **测试。** 每个技能都有自己的用例，来自复盘、记录的对话或手写，任意版本都能在这些用例上回测打分。
+- **安装。** 市场可以搜索 skills.sh 的开放索引，读取任意 GitHub 仓库、任何按 Agent Skills 发现协议发布技能的网站，以及技能目录；每次安装前都能预览，有安全审核的会一并显示。
+- **对话中。** 每个回合下方的一行列出这一回合用到的技能和复盘提出的建议，点一下就能打开。
+
+自进化默认关闭，在**插件 → 可组合记忆 → 技能管理**中开启。[技能管理指南](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/zh-CN/guides/skills.md)逐一介绍了每个页面。
+
 ## 快速开始
 
 第一次使用 DSH？[安装与启动](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/zh-CN/guides/installation.md)从一台空白电脑一直带你到第一条记忆，每一步都有截图。简要步骤如下，需要 [Node.js](https://nodejs.org/) 22.19 或更高：
@@ -114,6 +131,7 @@ Starter 固定一组经过测试、各自独立版本的包。同一时间只运
 | [dsh-mnemon-source-runtime](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-source-runtime/README.md) | 运行时记忆：USER.md、MEMORY.md、修订与本地存储 | 开启 |
 | [dsh-mnemon-source-documents](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-source-documents/README.md) | 项目档案：Markdown、检索、修订与归档 | 开启 |
 | [dsh-mnemon-source-memory-spaces](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-source-memory-spaces/README.md) | 记忆空间：长期证据及其 Provider | 开启 |
+| [dsh-mnemon-source-skills](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-source-skills/README.md) | 技能管理（实验性）：DSH 技能的版本、来自真实对话的自进化、测试与回测，以及汇集 skills.sh、GitHub 与 Agent Skills 网站的市场 | 开启（自进化默认关闭） |
 | [dsh-mnemon-strategy-default-three-tier](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-strategy-default-three-tier/README.md) | 分层策略：运行时记忆常驻，其余按需读取 | 选中 |
 | [dsh-mnemon-strategy-general](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-strategy-general/README.md) | 通用策略：全部可用来源共享一份预算 | 关闭 |
 | [dsh-mnemon-strategy-auto-capture](https://github.com/omdsh-dev/dsh-mnemon/blob/main/plugins/dsh-mnemon-strategy-auto-capture/README.md) | 主动记录：在回合中提示保留有用的事实 | 关闭 |
@@ -132,6 +150,7 @@ Starter 固定一组经过测试、各自独立版本的包。同一时间只运
 
 - 运行时记忆与项目档案是本地文件，Mnemon Native 也在本地；第三方 Provider 使用各自的服务与作用域。
 - 关闭组件不会删除其中的记忆，更换存储位置也不会搬移数据；需要迁移时使用 ZIP 备份。
+- 技能管理的进化数据（建议、测试用例以及复盘读过的对话，机密已替换）保存在本地存储根目录，不在备份之内。来自市场的技能会被 Agent 按原文执行，安装前请先预览。
 - Git 同步把同一份 Mnemon Pack 载荷以可读文件发布到你自己的仓库，另一台机器可以拉取；可以**立即同步**、在写入前**检查远端**，或按间隔**自动同步**。两台设备对同一主题写法不同时，两种写法都会保留，直到你接受并执行**记忆整理**给出的方案，见 [Git 仓库同步](https://github.com/omdsh-dev/dsh-mnemon/blob/main/docs/zh-CN/guides/operations.md#git-仓库同步)。
 - 已保存的 Provider 凭据只留在宿主上，不会被导出；但备份仍包含私有记忆，请妥善保护。
 - Source 与 Strategy 是受信任的进程内 JavaScript，**不是沙箱代码**；历史记忆永远不会凌驾于当前指令之上。

package/cordis.patch.yml

@@ -69,6 +69,10 @@
                 instanceId: byterover
               - use: dsh-mnemon-provider-supermemory
                 instanceId: supermemory
+        # Skills keep DSH's own files and their evolution data; the page lives
+        # in the Memory System and nothing joins a model View.
+        - id: mnemon-source-skills
+          name: dsh-mnemon-source-skills
         - id: mnemon-strategy-default-three-tier
           name: dsh-mnemon-strategy-default-three-tier
         # A second main Strategy; memoryView.strategyTypeId selects which one

package/lib/bundle.js

@@ -1,4 +1,4 @@
-import { t as prepareStarterResolution } from "./starter-resolution-DoFp9FLr.js";
+import { t as prepareStarterResolution } from "./starter-resolution-C2TDVxDr.js";
 //#region src/bundle.ts
 const groupKey = Symbol.for("cordis.group");
 const init = Symbol.for("cordis.init");