dsh-big-fat-fish-girl

DSH PLUGIN会话与消息

仓库 troodon/dsh-big-fat-fish-girl

会话与消息 L3 · 安装规范通过 安全审查:低风险 可信度:银 可安装插件
安装这个插件

dsh-big-fat-fish-girl 是 troodon 维护的 会话与消息 类 DSH 可安装插件, 仓库为 troodon/dsh-big-fat-fish-girl,许可证 Unlicense。 下面按「装什么、怎么装、验没验过、跑起来要注意什么」逐项列出。

一句话选型结论

按资料字段整理

dsh-big-fat-fish-girl 是会话与消息类的 DSH 可安装插件,验证等级 L3 · 安装规范通过,安全审查为低风险,有可直接复制的安装命令。验证通过不等于安全:插件会读写文件、执行命令或联网,装前先到仓库确认它做了什么;涉及密钥或系统权限的,建议先在隔离环境试装。

分类会话与消息 收录类型可安装插件 验证等级L3 · 安装规范通过 安全审查低风险
验证等级 L3 L3 · 安装规范通过
GitHub 星标 0 复刻 未标注 · 未关闭 issue 未标注
最近更新 2026-10-26 创建于 未标注
README 9.9 KB 正文体量,越大说明说明书写得越细

安装

可安装插件
在 DSH 里执行,装完用 dsh plugin list 确认已挂上;标了 L4 / L5 的命令已在沙盒里真实跑通
dsh plugin --profile web add github:troodon/dsh-big-fat-fish-girl#v0.8.0
  1. 1
    确认版本与运行环境

    仓库没有标注兼容版本与 Node.js 要求,先确认你的 DSH 版本能装,再继续。

  2. 2
    执行安装命令

    在 DSH 里执行上面的命令;多包仓库要注意区分具体子包名,别装错包。

  3. 3
    确认已挂载

    用 dsh plugin list 查看列表里是否出现这个插件,出现即挂载成功。

  4. 4
    试跑并检查权限

    加载后跑一个最小用例,确认它能正常工作;同时看它申请了哪些文件、命令或网络权限。

兼容 DSH 版本未标注版本不匹配时安装会报错
Node.js 要求未标注来自仓库 package.json 的 engines 字段
前置依赖无需要同时安装的同级依赖
npm 包名未发布到 npm可用 npm view 查看包信息

验证等级与安全审查

L3 · 安装规范通过
自动化流程结论,不是人工审核
验证等级L3 · 安装规范通过安装命令可解析、DSH 版本已声明,未做实际安装
安全审查低风险静态分析结论,未标注不代表已通过审查
仓库健康度活跃按仓库活跃度与最近更新时间给出
可信度银综合仓库活跃度与验证结果

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 验证等级 L3:安装命令能解析、兼容版本已声明,至少不会一装就报格式错。
  • 静态安全审查为低风险,没有发现明显可疑行为。
  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有声明 Node.js 版本要求,版本不匹配可能装不上。
  • 验证通过不等于安全:插件会读写文件、执行命令或联网,装前先到仓库确认它做了什么。

更适合

  • 需要在会话与消息场景里给 DSH 补能力的人。
  • 能接受在本地装依赖、愿意自己确认权限的人。

不太适合

  • 只使用 DSH 官方内置能力、不接受第三方插件的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
插件验证等级安全审查星标兼容版本安装命令
dsh-big-fat-fish-girl当前 L3 · 安装规范通过 低风险 0 未标注 有
dsh-desktop L3 · 安装规范通过 低风险 3.0 万 未标注 无
openpets L4 · 沙盒安装通过 低风险 1,274 未标注 有
deeptide L4 · 沙盒安装通过 低风险 1,104 0.1.1-rc.2 有
jingyun-dsh L3 · 安装规范通过 低风险 1,004 未标注 有
study-mate 未做沙盒验证 未标注 856 未标注 无

仓库档案

会话与消息
字段来自项目仓库与公开索引,中文说明由本站标注
作者 / 维护方troodon仓库 dsh-big-fat-fish-girl · 默认分支 未标注
分类会话与消息desktop
收录类型可安装插件声明了 dsh.bundle、可用 dsh plugin add 安装的条目
开发语言未标注仓库主语言
许可证Unlicense商用或二次分发前请确认条款原文
星标 / 复刻 / issue0 / 未标注 / 未标注收录时的数值,非实时
创建 / 更新未标注 · 2026-10-26最后推送 未标注
本站收录2026-10-30首次进入本站目录的时间

README 全文

9.9 KB
作者原文,中文版优先;只做安全转义,不改内容

Big Fat Fish Girl for DeepSeek Harness



Big Fat Fish Girl is a desktop pet for DeepSeek Harness. She stays above other apps in a
transparent window while Harness is minimized and follows work across all chats. She shows
current activity, tool use, requests, completion and errors.

Clicking her brings Harness forward on its last-open chat, not a specific session.

Features

  • Tool activity: scenes for reading, editing, commands and general work, with distinct
    commentary for browsing and search. Tool names and arguments are not sent to the browser.
  • Requests: pending requests take priority over other animations.
  • Token usage: displays provider-reported input and output counts after completion when
    available. It does not estimate costs or balances.
  • Idle behavior: cursor gaze, hover reactions, random animations, sleep and occasional
    horizontal movement. Each behavior can be disabled in Harness settings.
  • Long runs: an optional snack animation can appear after 3–5 minutes of continuous work.
  • Languages: English by default, with Russian and Simplified Chinese available in settings.
  • Character: DeepSeek Chan (Big Fat Fish Girl) is the only bundled skin.
  • Reduced motion: uses a stable frame for each state and disables transitions.

Generated from every shipped DeepSeek Chan atlas by node scripts/build-preview.mjs.

Requirements

  • DeepSeek Harness 0.2.0-rc.2 (the event contract is verified against it — see
    docs/HARNESS-CONTRACT-v0.2.0-rc.2.md).
  • A Harness profile with the web interface, such as web or desktop.
  • Runtime dependency: @deepseek-ai/schemastery.

See it move

These captioned loops are rendered from the shipped DeepSeek Chan art by
node scripts/build-preview.mjs; the atlas overview above includes every animation sheet.

What she says

These samples were captured from the running preview with npm run preview:lines.

English — default:

Russian:

Simplified Chinese:

The amber bubble is Harness's own approval colour, used when she needs you or is changing
something. A quiet moment has no line at all — the last frame above.

Install

Download the plugin .tgz and the separate macOS app from the same
GitHub release. You can also build
the app from that release's source tag.

The plugin package contains runtime code and sprite atlases, without demo GIFs, screenshots,
source art or build tools. Install the .tgz, not GitHub's larger Source code archive.
Follow the macOS install guide
for both routes, checksums, Gatekeeper and post-install verification.

Settings

Settings live in Harness's Settings panel, in the Big Fat Fish Girl section
(size and opacity are sliders — the pet changes as you drag); the same values are also
editable from the Plugins panel's Configure form. The pet carries no menu or panel of its
own, and the position is the only thing kept per window.

Setting What it does
Size / Opacity 88–172 px and 0.35–1
Reduce motion Freezes to each state's readable anchor frame and stops transitions
React to hover On/off for the cursor reactions
Random antics On/off for the spontaneous idle actions
Snacks On/off for the snack after long work
Token count display On/off for provider-reported counts after a completed job
Wandering On/off for the autonomous strolls
Language English by default; Russian and Simplified Chinese are also available

For state aggregation and priority details, see docs/ARCHITECTURE.md.

Desktop shell

For shell startup and bridge details, see desktop/README.md. Build and
platform information is in desktop/tauri/README.md.

Development

Node.js 22 or newer is required for development. Install dependencies and run the checks:

npm install
npm run check
npm run build
npm run build:desktop
node preview/server.mjs

npm run preview:art regenerates the art previews. Install its pinned Pillow version with
python3 -m pip install --requirement scripts/preview-requirements.txt. See preview/README.md for
the local playground, docs/ARCHITECTURE.md for the module map, and
docs/adding-an-animation.md to add an animation.

Editable PNGs and original pose boards live on the separate art-source branch. Normal builds,
tests and previews use the committed runtime atlases and do not download source art. See the
art source guide only if you want to rebuild the atlases.

Known limitations

  • The tested macOS desktop build is arm64; Intel is untested.
  • Windows builds are untested. Linux builds are not available.
  • The pet stays above ordinary windows; full-screen coverage is not promised.
  • Settings are per profile. Installing the plugin into two profiles gives you two
    independent pets; the pet's window position is the only thing kept per browser window.
  • Token counts are provider-reported or omitted. The plugin does not estimate money;
    cost, balance and remaining-quota APIs are not part of the verified Harness contract.
  • Other Harness versions are not certified. The event contract is verified against
    0.2.0-rc.2. The 0.2.1 startup-origin fix does not verify that version's full event contract.

Privacy

  • The plugin makes no outbound requests and sends no telemetry.
  • It keeps transient session IDs and status changes in memory. It never reads prompt text,
    responses, tool arguments or session titles.
  • For navigation, the plugin sends an opaque ID for the session with an open request or task.
    The ID contains no content, and the host accepts it only while that session is open.
  • Provider-reported input and output token counts are summed per task and sent as numbers
    only; model names, message IDs and content are not sent with them.
  • Display preferences stay in the Harness profile; only the pet's window position is stored
    in browser local storage.
  • See the verified Harness contract for event details.

Issues and feature requests: GitHub Issues.

License

Code: Unlicense, dedicated to the public domain.
Character artwork: not covered by the code dedication; reuse terms are not granted here — see ASSET_LICENSE.md.

常见问题

6 问
答案由本站依据条目字段整理,不引用上游原文
dsh-big-fat-fish-girl 怎么安装?

在 DSH 里执行 `dsh plugin --profile web add github:troodon/dsh-big-fat-fish-girl#v0.8.0` 即可。

dsh-big-fat-fish-girl 的验证等级是什么意思?

L3 · 安装规范通过。安装命令可解析、DSH 版本已声明,未做实际安装

dsh-big-fat-fish-girl 安全吗?

低风险。

dsh-big-fat-fish-girl 装之前要先确认什么?

先看它会读写哪些文件、执行哪些命令、连接哪些网络地址;验证等级只说明装得上、跑得起来,不等于安全。涉及密钥或系统权限的,建议先在隔离环境试装。

dsh-big-fat-fish-girl 用的是什么许可证?

上游标注为 Unlicense。商用或二次分发前请到仓库确认许可证原文。

这条记录的数据可靠吗?

本站按仓库名去重后合并整理,字段以作者仓库与 npm 元数据为准;安装命令、仓库地址与 README 保持原文,带 L4 / L5 标记的安装命令已在隔离沙盒里真实跑通。

会话与消息
按星标排序,可作替代方案

dsh-desktop

dsh-desktop

为 DeepSeek Harness (DSH) 插件生态打造的现代化桌面端解决方案。万物皆「插件」,桌面本身也是「插件」。

L3 · 安装规范通过 3.0 万 星

openpets

openpets

L4 · 沙盒安装通过 1,274 星

deeptide

deeptide

L4 · 沙盒安装通过 1,104 星

jingyun-dsh

jingyun-dsh

井云 DSH 客户端插件,为 DeepSeek Harness 注入会员登录、订阅支付、算力管控与应用市场,让智能体和技能快速变成可交易商品

L3 · 安装规范通过 1,004 星

study-mate

study-mate

StudyMate 是 DSH 的「学习模式」预设与技能包,为数学、计算机科目定制学习路线、讲解课件与练习评估,跨科目共享记忆。

未标注验证等级 856 星

插件的元信息来自作者仓库;分类、验证等级与安全结论的中文说明为中文整理,最近核验 —。