@morehao/dsh-skills-console 0.5.1 → 0.5.3

package/README.en.md

@@ -0,0 +1,170 @@
+<p align="center"><img src="icon.svg" width="64" height="64" alt="Skills Console icon" /></p>
+
+# Skills Console · dsh-skills-console
+
+[简体中文](README.md) · **English**
+
+[![npm version](https://img.shields.io/npm/v/@morehao/dsh-skills-console.svg)](https://www.npmjs.com/package/@morehao/dsh-skills-console) [![check](https://github.com/morehao/dsh-skills-console/actions/workflows/check.yml/badge.svg)](https://github.com/morehao/dsh-skills-console/actions/workflows/check.yml) [![license](https://img.shields.io/npm/l/@morehao/dsh-skills-console.svg)](LICENSE)
+
+The skills console for DSH (DeepSeek Harness): **install, update, remove, toggle and share skills with other agents — plus a subscription list**.
+
+It gathers the skills on your machine into one list: what you have, where it came from, who can use it, and whether it is up to date. Turn on what you use, turn off what you don't; when another agent needs one, link it over with a click — the skill itself always stays in one place.
+
+## Why you need it
+
+- **One place for every skill**: installed skills and skills already on the machine live in the same list — source, version, shared agents and update time at a glance, no digging through directories.
+- **Toggle on demand, save context**: disable a skill you don't need for now and DSH's model invocations and `/` manual invocations stop seeing it; turn it back on and it is back. Skills on disk, sharing, and other agents are untouched.
+- **One skill, many agents**: most agents (OpenCode, Cursor, Gemini CLI, …) read the shared skills directory directly and work as soon as a skill is installed; ones that only read their own directory (such as Claude Code) can be linked with a single checkbox in the panel.
+- **Installing is not the end**: upstream repos keep moving. The console checks for new commits, first tells you which skills are affected, and then you decide whether to update.
+- **A wrong delete is recoverable**: removals go to the trash for 7 days; a failed update rolls back to the previous version.
+
+## Features
+
+| Capability | Description |
+| --- | --- |
+| Inventory | Collects the skills on this machine — source, version, visibility and shared agents in one view; skills already in the shared directory are adopted automatically the first time you open it |
+| Install | Enter `owner/repo` or a GitHub URL; first detects which skills the repo has and what each does; install everything or pick a few |
+| Update | Checks upstream commits, lists the commit range and affected skills, updates in one click after you confirm; locally modified skills ask for confirmation first |
+| Remove | Pre-flight check (never touches what others created), then a snapshot into the trash (kept 7 days) |
+| Toggle | One-click on/off; once disabled, neither DSH's model invocations nor `/` manual invocations see it; other agents and files on disk are unaffected |
+| Share | Links only agents that read their own directory; nothing is shared by default; agents that read the shared directory directly are not listed |
+| View | Open a skill's body and files read-only: rendered / raw toggle, directory tree, single-file preview, plus source, hash and sharing info |
+| AI intro | A ✨ per row opens a model-written summary of what the skill does and when to use it; only `SKILL.md` and the file-name list are sent, results are cached locally, and unchanged content is not regenerated |
+| Subscriptions | Follow upstream changes of an `owner/repo`; snapshots only, no skill download. Expand an entry to see the author's own description, or let the model write an AI intro before deciding. ref / subdirectory optional; an empty ref follows the default branch |
+| Scope | Manages user-level shared skills; per-project toggling is not supported yet. When a same-named copy exists in a project or user directory, that copy still wins visibility |
+
+## Screenshots
+
+Screenshots are from a real DSH Web session on v0.5.2 (dark theme). All assets live in [`assets/`](https://github.com/morehao/dsh-skills-console/tree/main/assets) (bitmap icons and UI screenshots) and are not published to npm, so they are referenced by absolute URL and render on both GitHub and the npm page.
+
+**Installed list**: source, invocation policy, shared agents and update status in one screen; the top switches between "Installed" and "Subscriptions".
+
+![Installed list: source, invocation policy, shared agents and update status in one screen](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-1-installed.png)
+
+**Read-only viewer**: the list stays put while the right-hand drawer renders `SKILL.md`; switch to raw, browse the file tree and info, or open a full page.
+
+![Read-only viewer: the right drawer renders SKILL.md, with raw text and the file tree](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-2-viewer.png)
+
+**AI intro**: click the ✨ in a row for a model summary of what the skill can do and when to use it; it opens as a centered dialog like the other modals, and once generated it reads from the local cache.
+
+![AI intro: a centered dialog shows the model-generated capability summary](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-3-intro.png)
+
+**Subscriptions**: only upstream snapshots of `owner/repo`; expand an entry to see how many skills the repo has, which are installed and their status.
+
+![Subscriptions: expand a repo to see its skills and install status](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-4-subscriptions.png)
+
+**Add subscription**: enter `owner/repo` or a GitHub URL; ref and subdirectory are optional; follow only, no skill download.
+
+![Add subscription: fill in source, ref and subdirectory](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-5-subscribe-dialog.png)
+
+## Interface
+
+- **Two entry points**: by default it lives in the "Skills Console" section of DSH settings; turn on "Sidebar entry" in the header ⚙ and a standalone panel appears in the left sidebar, effective immediately.
+- **One page, two tabs**: Installed / Subscriptions. The header holds "Auto-share on install" and "Add skill / Add subscription" on the right.
+- **Five actions per row**: toggle, view, update, share settings, delete; the ✨ next to a skill name opens the AI intro; "View" is hidden on narrow panels, and the skill name itself remains the entry point. Status labels cover up to date / update available / locally modified / removed upstream / broken link / no upstream.
+- **No layout jumps on narrow panels**: the action group wraps naturally with panel width and the plugin name always stays whole; the list switches to two columns by width; scrolling happens only in the list area, so switching tabs or resizing never flips the whole page.
+- **Many agents don't crowd**: at most 3 are shown per row and the rest fold into `+n`; hover or focus reveals all.
+- **Findable**: search by name / source / agent, filter by "updates / not checked / disabled", view flat or grouped by agent.
+- **Not checked vs. no upstream**: "Not checked" only means no upstream verdict was obtained this time (rate limit, network, repo 404, etc.) and will be retried next time; skills without a usable source are marked "No upstream" and skipped in update checks.
+
+## Viewing skill content
+
+While managing skills you may want to know "what does this skill say, and is it worth keeping" — click the skill name or "View" on the row:
+
+- **Drawer quick look**: the list stays put and a read-only drawer slides in from the right; close it with Esc / the overlay / ×, and you return to the previous filter and scroll position.
+- **Full page for a deep read**: "Open as page" at the bottom of the drawer switches to a full page (body + contents on the left), with a top bar to go back to the list.
+- **Three tabs**: Content (rendered / raw toggle; raw keeps the frontmatter), Files (directory tree + single-file preview; `.md` renders by default and can switch to raw, other text shows its source directly), Info (source, ref, commit, skillPath, content hash, install / update time, shared agents, enabled state, local changes).
+- **Read-only**: no inputs and no save button. Update / share settings are enabled depending on writability; when the state file is corrupt you can still view, but management actions are disabled.
+- **What can be previewed**: text files that decode as UTF-8 (`.md` / `.html` / `.jsx` / `.css` / `.sh` / extension-less scripts…) show their source directly; the per-file cap is 256 KB (longer content is truncated with a notice); images, audio/video, archives and other binaries show "Preview not supported" with the file name / size.
+
+## AI intros (model-generated)
+
+Some skills only have an English description, and some titles alone don't tell you what they do. Every installed row has a `✨ Intro` next to the skill name:
+
+- **Open and read**: generated intros load from the local cache and open instantly; only when there is none does a "Generate intro" button appear, and only clicking it calls the model.
+- **What it writes**: a 2–3 sentence capability summary — what it does, when to use it, and **how to trigger it**: trigger words / example requests copied from the `SKILL.md` when present, otherwise 2–4 keywords (or one example request) users can literally say, or the invocation itself (skill name / slash command); key limitations when needed; in the model's own words, not a word-for-word translation.
+- **What it sends**: the `SKILL.md` body (truncated beyond `maxInputBytes`) and the list of file names in the skill directory (up to 30); **no other file contents are read**; the language follows the UI (Chinese / English).
+- **Cached locally**: `$DSH_HOME/skills-console/intros.json`, keyed by skill name + language; when the `SKILL.md` content hash changes it warns "may be out of date" and only regenerating overwrites it. Failures are not cached.
+- **Default model**: `config.intro.provider/model` when configured, otherwise DSH's current default model; only one generation runs at a time, and repeated clicks return "already generating".
+- **Subscription candidates too**: expanded subscriptions have `✨ Intro` on each entry; it fetches the remote `SKILL.md` at the snapshot `commit` and generates from that, with the commit in the cache key, and it never blocks "Install". Expanding also shows the author's own frontmatter description for free.
+- **How to turn it off**: `config.intro.enabled = false` (hides the ✨ entry). Other knobs are under "Data and configuration".
+
+## Installation
+
+The plugin ships its own download and extraction code: **no Node / npm / npx / git required beforehand**; the first install needs network access to fetch dependencies.
+
+Install a local link inside Harness:
+
+```bash
+plugin_manager install_bundle(target = <absolute path to this repo>)
+```
+
+Or add it to the profile's `cordis.patch.yml`:
+
+```yaml
+- insert:
+    - id: dsh-skills-console
+      name: '@morehao/dsh-skills-console'
+```
+
+The panel shows up in DSH Web's settings page / sidebar.
+
+## Quick start
+
+1. **Add skill**: click "Add skill" in the top right, enter `owner/repo` or a GitHub URL, then click "Detect"; when the repo has several skills, tick the ones to install (multi-select, all selected by default, each with its description), and a single skill is selected automatically. If nothing is detected you can type skill names by hand; leaving them empty installs everything in the repo.
+2. **Share with other agents**: only needed for agents that read their own directory (for example Claude Code). Click "Share settings" on a row and tick the target agents; agents that already read the shared directory are not listed. To choose which agents new installs are shared with by default, use "Auto-share on install" in the header (nothing is ticked by default).
+3. **Check for updates**: click the refresh icon in the list header; rows with updates light up "Update", or update everything at once with "Update all". Locally modified skills ask for confirmation first.
+4. **Invocation policy**: click the "Auto / Manual only / Fully disabled" status button on a row and switch in the centered invocation-policy dialog. Auto = both the model and `/` can invoke it; Manual only = only the `disable-model-invocation` frontmatter key in `SKILL.md` changes and `/` still works; Fully disabled = `SKILL.md` is renamed to `SKILL.md.disabled` with a marker left behind, and restoring returns to the mode from before. When the author already declared manual-only in `SKILL.md`, the chip is read-only and is never overridden. While parked, install and update are rejected so the file cannot be silently written back.
+
+## Safety and reliability
+
+- **Only touch what is ours**: the plugin maintains only the links it created; it runs a pre-flight check before deleting and aborts if a link to remove has become a real directory or a foreign link; same-named links created by others are never touched or removed.
+- **Where file changes stop**: Auto only writes plugin state; Manual only changes one frontmatter line, with a backup first; Fully disabled only renames + writes a marker, leaving the content byte-for-byte identical. The v3 lock, shared links and the body are never touched; updates replay the invocation policy and still ask for confirmation for other local changes.
+- **Rollback included**: removals go to the trash (7 days); every update is backed up first, restores the previous version on failure, and can be rolled back manually when needed.
+- **Read-only when state is bad**: if the state file is corrupt or its version is unknown, the plugin degrades to read-only, management actions are disabled, and it neither guesses the structure nor auto-repairs.
+- **No arbitrary paths**: targets in requests are always resolved through the registry; shared links only ever point to the same-named skill under the shared directory.
+- **Big repos still update**: large repos with design assets (50 MB+ archives) are not blocked by the size cap.
+- **"Manual only" has a real entry point**: it can be switched in the console, or declared by the skill author in `SKILL.md`; an author declaration is shown read-only, never overridden, and never treated as an error.
+- **Same-named copies are called out**: when a project or user directory holds another copy of a skill, the UI says outright that "the model sees a different copy" instead of pretending the visible one is the shared-directory copy.
+- **Rate limits have a fallback**: public repos work anonymously; when the GitHub quota runs out it switches channels automatically to keep checking, and you can set `GITHUB_TOKEN` to raise the quota from 60 to 5000 requests per hour — private repo installs rely on it too.
+
+## Data and configuration
+
+| Item | Location |
+| --- | --- |
+| Skill itself | `~/.agents/skills` (the skills directory shared by DSH and most agents) |
+| Plugin state | `$DSH_HOME/skills-console/state.json` (`DSH_SKILLS_CONSOLE_HOME` changes the root) |
+| Backups / trash | `$DSH_HOME/skills-console/{backups,trash}`; originals rewritten by Manual only are kept separately in `backups/invocation/` |
+| AI intro cache | `$DSH_HOME/skills-console/intros.json` (installed skills by name, subscription candidates by `source@commit:path`; only the generated text and metadata, never the `SKILL.md` body) |
+| Event log | `$DSH_HOME/skills-console/dsh-skills-console.log` (JSONL, rotates at 1 MB) |
+
+The skill itself stays in the shared directory; sharing only creates links and never moves files.
+
+AI intro settings (in the plugin `config` of the profile / `cordis.patch.yml`):
+
+| Key | Default | Description |
+| --- | --- | --- |
+| `intro.enabled` | `true` | `false` hides the ✨ entry and the API returns 501 |
+| `intro.provider` / `intro.model` | empty | empty uses DSH's current default model; both must be set together |
+| `intro.maxInputBytes` | `24576` | cap on the `SKILL.md` sent to the model (8–48 KB); longer input is truncated and marked in the prompt |
+| `intro.maxOutputTokens` | `300` | output cap (96–1024) |
+| `intro.timeoutMs` | `30000` | per-generation timeout (5–60 s); timeouts return 504 |
+
+## Development and publishing
+
+Source layout, test strategy and local `link:` verification are in [docs/development.md](docs/development.md); visibility, data placement and implementation boundaries in [docs/internals.md](docs/internals.md); the release process and troubleshooting checklist in [docs/publishing.md](docs/publishing.md). AI agents working in this repo should read [AGENTS.md](AGENTS.md) first.
+
+## Boundaries (explicitly out of scope)
+
+- No skill marketplace / search, and no taking over or rewriting other tools' own agent distribution.
+- No editing skill bodies: the viewer is read-only; only "Manual only" touches the `disable-model-invocation` frontmatter key, and "Fully disabled" only renames files — nothing else.
+- No file-level diff preview: only commit ranges and affected skills; to inspect individual lines, run `git diff <from>..<to> -- <skillDir>` in a terminal.
+- No image / binary preview.
+- No cross-machine sync, and no token is ever written to disk.
+- No process-level hiding: toggling in Auto only affects DSH; Fully disabled renames `SKILL.md`, so other agents that read that directory can't see it either, but it does not prevent external writes from putting the file back.
+- No project-level toggling: same-named skills in projects stay visible per DSH's precedence.
+- No taking over the host preset layer or skill sources bundled with other plugins; diagnostics mark them as blind spots rather than pretending to support them.
+
+## License
+
+MIT

package/README.md

@@ -1,10 +1,10 @@
-<img src="icon.svg" width="64" height="64" alt="技能控制台图标" align="right" />
+<p align="center"><img src="icon.svg" width="64" height="64" alt="技能控制台图标" /></p>
 
 # 技能控制台 · dsh-skills-console
 
-[![npm version](https://img.shields.io/npm/v/@morehao/dsh-skills-console.svg)](https://www.npmjs.com/package/@morehao/dsh-skills-console)
-[![check](https://github.com/morehao/dsh-skills-console/actions/workflows/check.yml/badge.svg)](https://github.com/morehao/dsh-skills-console/actions/workflows/check.yml)
-[![license](https://img.shields.io/npm/l/@morehao/dsh-skills-console.svg)](LICENSE)
+**简体中文** · [English](README.en.md)
+
+[![npm version](https://img.shields.io/npm/v/@morehao/dsh-skills-console.svg)](https://www.npmjs.com/package/@morehao/dsh-skills-console) [![check](https://github.com/morehao/dsh-skills-console/actions/workflows/check.yml/badge.svg)](https://github.com/morehao/dsh-skills-console/actions/workflows/check.yml) [![license](https://img.shields.io/npm/l/@morehao/dsh-skills-console.svg)](LICENSE)
 
 DSH（DeepSeek Harness）的技能控制台：**装、更新、删除、启停、共享至其他 Agent，外加一份订阅清单**。
 
@@ -29,15 +29,39 @@ DSH（DeepSeek Harness）的技能控制台：**装、更新、删除、启停
 | 启停 | 一键开关；停用后 DSH 的模型调用与 `/` 手动调用都看不到它，其他 Agent 与磁盘文件不受影响 |
 | 共享 | 只给「只读自己目录」的 Agent 建链，默认不共享；直接读共享目录的 Agent 不出现在列表里 |
 | 查看 | 只读打开技能正文与文件：渲染 / 原文切换、目录树、单文件预览，外加来源、哈希与共享信息 |
-| 技能介绍 | 列表每行一个 ✨，点开就能看到模型写的「这个技能能做什么、什么时候用」；只发 `SKILL.md` 与文件名清单，结果缓存在本地，内容没变不重复调用 |
-| 订阅 | 关注 `owner/repo` 的上游变化，只记快照、不下载技能；展开条目能看到作者原描述，也可以让模型写一段 AI 介绍再决定装不装；ref / 子目录可选，不填 ref 时跟随默认分支 |
+| 技能简介 | 列表每行一个 ✨，点开就能看到模型写的「这个技能能做什么、什么时候用」；只发 `SKILL.md` 与文件名清单，结果缓存在本地，内容没变不重复调用 |
+| 订阅 | 关注 `owner/repo` 的上游变化，只记快照、不下载技能；展开条目能看到作者原描述，也可以让模型写一段 AI 简介再决定装不装；ref / 子目录可选，不填 ref 时跟随默认分支 |
 | 作用范围 | 管 user 级共享技能；暂不支持按项目启停。项目或用户目录里有同名副本时，副本仍优先可见 |
 
+## 截图
+
+截图取自 v0.5.2 的 DSH Web 实机（深色主题）。素材都放在 [`assets/`](https://github.com/morehao/dsh-skills-console/tree/main/assets)（位图图标与界面截图），不随 npm 包发布，所以这里用绝对地址引用，GitHub 与 npm 页面都能看到。
+
+**已安装清单**：来源、调用策略、共享的 Agent、更新状态一屏看全；顶部切「已安装 / 订阅」。
+
+![已安装清单：来源、调用策略、共享的 Agent 与更新状态一屏看全](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-1-installed.png)
+
+**只读查看器**：列表不动，右侧抽屉渲染 `SKILL.md`；可切原文、看文件树与信息，也能进整页阅读。
+
+![只读查看器：右侧抽屉渲染 SKILL.md，可切原文与文件树](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-2-viewer.png)
+
+**AI 简介**：点行内 ✨，看模型概括的「能做什么、什么时候用」；和调用策略等弹窗一样居中打开，生成过就直接读本地缓存。
+
+![AI 简介：居中弹窗展示模型生成的能力简介](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-3-intro.png)
+
+**订阅清单**：只记 `owner/repo` 的上游快照；展开看仓库里有几个技能、哪些已装、各自状态。
+
+![订阅清单：展开仓库查看技能与安装状态](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-4-subscriptions.png)
+
+**添加订阅**：填 `owner/repo` 或 GitHub URL，ref 与子目录可空；只关注，不下载技能。
+
+![添加订阅：填写来源、ref 与子目录](https://raw.githubusercontent.com/morehao/dsh-skills-console/main/assets/screenshot-5-subscribe-dialog.png)
+
 ## 界面
 
 - **两个入口**：默认在 DSH 设置页的「技能控制台」分区；页头 ⚙ 里打开「侧边栏入口」，左侧栏就会出现独立面板，切换立即生效。
 - **一页两 tab**：已安装 / 订阅。页头右侧是「新安装时自动共享」与「添加技能 / 添加订阅」。
-- **每行五个动作**：开关（启停）、查看、更新、共享设置、删除；技能名旁的 ✨ 打开 AI 介绍；窄面板下隐藏「查看」，技能名本身仍是入口。状态标签覆盖最新 / 可更新 / 本地已改 / 上游已移除 / 软链断裂 / 无上游等。
+- **每行五个动作**：开关（启停）、查看、更新、共享设置、删除；技能名旁的 ✨ 打开 AI 简介；窄面板下隐藏「查看」，技能名本身仍是入口。状态标签覆盖最新 / 可更新 / 本地已改 / 上游已移除 / 软链断裂 / 无上游等。
 - **窄面板不跳版**：动作组按面板宽度自然换行，插件名始终完整；列表按宽度切两列；滚动只发生在列表区，切 tab、调整窗口都不会让整页翻面。
 - **Agent 多了不挤**：每行最多展示 3 个，其余折成 `+n`，悬浮或聚焦即可看全。
 - **找得到**：支持按名称 / 来源 / Agent 搜索，按「可更新 / 待检查 / 已停用」筛选，平铺或按 Agent 分组。
@@ -53,16 +77,16 @@ DSH（DeepSeek Harness）的技能控制台：**装、更新、删除、启停
 - **只读**：没有输入框和保存按钮。更新 / 共享设置按可写状态决定是否可用；状态文件损坏时仍可查看，管理动作禁用。
 - **能预览什么**：能按 UTF-8 解码的文本文件（`.md` / `.html` / `.jsx` / `.css` / `.sh` / 无扩展名脚本…）直接显示源码；单文件上限 256 KB（超出截断并提示）；图片、音视频、压缩包等二进制显示「暂不支持预览」和文件名 / 大小。
 
-## 技能介绍（AI 生成）
+## 技能简介（AI 生成）
 
-有些技能只有英文说明，也有的光看标题不知道能干什么。已安装列表里每行技能名旁都有一个 `✨ 介绍`：
+有些技能只有英文说明，也有的光看标题不知道能干什么。已安装列表里每行技能名旁都有一个 `✨ 简介`：
 
-- **点开即看**：生成过的直接显示本地缓存，秒开；没生成过才出现「生成介绍」按钮，点它才调模型。
-- **写什么**：2–3 句能力介绍——能做什么、适合什么场景 / 典型触发，必要时点出关键限制；用模型自己的话概括，不是逐字翻译。
+- **点开即看**：生成过的直接显示本地缓存，秒开；没生成过才出现「生成简介」按钮，点它才调模型。
+- **写什么**：2–3 句能力简介——能做什么、什么时候用，并写清**怎么触发**：优先照抄 `SKILL.md` 里的触发词 / 示例请求，原文没写就按作者描述归纳成可以直接说的话（关键词或一句示例），也可以写技能名 / 斜杠命令；必要时点出关键限制；用模型自己的话概括，不是逐字翻译。
 - **发什么**：`SKILL.md` 正文（超出 `maxInputBytes` 截断）与技能目录的文件名清单（最多 30 个），**不读其他文件内容**；语言跟随界面（中文 / English）。
 - **缓存在本地**：`$DSH_HOME/skills-console/intros.json`，按技能名 + 语言分开存；`SKILL.md` 内容哈希变了会提示「可能过期」，重新生成才覆盖。失败不缓存。
 - **默认模型**：优先用 `config.intro.provider/model`，没配就用 DSH 当前默认模型；同一时间只跑一个生成，重复点击返回「正在生成」。
-- **订阅候选也能看**：展开订阅后，条目行同样有 `✨ 介绍`；按快照 `commit` 拉远端 `SKILL.md` 再生成，缓存 key 带 commit，不挡「安装」。展开时还会显示作者自己的 frontmatter 描述，不花额度。
+- **订阅候选也能看**：展开订阅后，条目行同样有 `✨ 简介`；按快照 `commit` 拉远端 `SKILL.md` 再生成，缓存 key 带 commit，不挡「安装」。展开时还会显示作者自己的 frontmatter 描述，不花额度。
 - **关闭方式**：`config.intro.enabled = false`（界面隐藏 ✨ 入口）。其他可调项见「数据与配置」。
 
 ## 安装
@@ -90,7 +114,7 @@ plugin_manager install_bundle(target = <本仓库绝对路径>)
 1. **添加技能**：点右上角「添加技能」，填 `owner/repo` 或 GitHub URL，点「检测技能」；仓库里有多个技能时勾选要装的（支持多选，默认全选，列表带每个技能的说明），只有一个会默认勾上。检测不到时可以手填技能名，留空则装仓库内全部。
 2. **共享至其他 Agent**：只有「只读自己目录」的 Agent 才需要这一步（例如 Claude Code）。点行内「共享设置」勾选目标 Agent；本来就读共享目录的 Agent 不在列表里。之后新装技能默认共享至哪些 Agent，用页头「新安装时自动共享」设定（默认一个都不勾）。
 3. **检查更新**：点列表头上的刷新图标；有更新的行会亮起「更新」，也可以一次「全部更新」。本地改过的技能会先要求确认。
-4. **调用策略**：点行内「自动 / 仅手动 / 完全停用」状态按钮切换，弹层里同时给出各 Agent 的覆盖情况。自动＝模型与 `/` 都能调用；仅手动＝只改 `SKILL.md` 的 `disable-model-invocation`，`/` 仍可用；完全停用＝把 `SKILL.md` 改名为 `SKILL.md.disabled` 并留 marker，恢复时回到停用前的模式。作者已在 `SKILL.md` 里声明仅手动时，chip 只读展示，不会被覆盖。停用期间更新 / 安装会被拒绝，避免文件被悄悄写回。
+4. **调用策略**：点行内「自动 / 仅手动 / 完全停用」状态按钮，在居中的调用策略弹窗里切换。自动＝模型与 `/` 都能调用；仅手动＝只改 `SKILL.md` 的 `disable-model-invocation`，`/` 仍可用；完全停用＝把 `SKILL.md` 改名为 `SKILL.md.disabled` 并留 marker，恢复时回到停用前的模式。作者已在 `SKILL.md` 里声明仅手动时，chip 只读展示，不会被覆盖。停用期间更新 / 安装会被拒绝，避免文件被悄悄写回。
 
 ## 安全与可靠性
 
@@ -111,12 +135,12 @@ plugin_manager install_bundle(target = <本仓库绝对路径>)
 | 技能本体 | `~/.agents/skills`（DSH 与多数 Agent 共用的技能目录） |
 | 插件状态 | `$DSH_HOME/skills-console/state.json`（`DSH_SKILLS_CONSOLE_HOME` 可换根目录） |
 | 备份 / 回收站 | `$DSH_HOME/skills-console/{backups,trash}`；仅手动改写的原文另存 `backups/invocation/` |
-| 技能介绍缓存 | `$DSH_HOME/skills-console/intros.json`（已安装技能按名字、订阅候选按 `source@commit:path`；只存生成的文本与元数据，不存 `SKILL.md` 原文） |
+| 技能简介缓存 | `$DSH_HOME/skills-console/intros.json`（已安装技能按名字、订阅候选按 `source@commit:path`；只存生成的文本与元数据，不存 `SKILL.md` 原文） |
 | 事件日志 | `$DSH_HOME/skills-console/dsh-skills-console.log`（JSONL，1MB 滚动） |
 
 技能本体留在共享目录，共享只是建链，不搬文件。
 
-技能介绍的可调项（写在 profile / `cordis.patch.yml` 的插件 `config` 里）：
+技能简介的可调项（写在 profile / `cordis.patch.yml` 的插件 `config` 里）：
 
 | 键 | 默认 | 说明 |
 | --- | --- | --- |