Riffkit 可安装的 DSH 包 - riff 获胜短视频至您的 DeepSeek Harness 代理。
Riffkit as an installable DSH bundle — riff a winning short video into your own, from your DeepSeek Harness agent.
安装
dsh plugin --profile web add github:riffkit/dsh-pluginGitHub 源码安装:首次需按提示配置 allowBuilds 构建授权后重试
安装与环境配置指引、插件开发教程见 DSH 中文社区文档 ↗
安装即在你的机器上以你的权限运行第三方代码——它可读写文件、使用凭据、访问网络,DSH 的工具审批不会为插件代码加沙箱。「检测到 manifest」仅代表发现 dsh.bundle / dsh.plugin 清单,不构成兼容性或安全审查;安装前请审阅源码,不熟悉的插件先在不含密钥的环境试用。
README
The Riffkit agent skill, packaged as an installable DSH bundle.
dsh plugin add @riffkit/dsh-plugin
Riff a winning short video into your own: Riffkit studies why a video worked — the hook, the pacing, the emotional beats — and rebuilds that formula as new footage with your product, your character and your language. The source clip is never re-uploaded; what carries over is the structure, not a single frame.
What this package is
A packaging layer, nothing more. It ships one thing — the skill — and registers
its own directory as an additional skill root through cordis.patch.yml:
- insert:
- id: riffkit-skill-root
name: '@deepseek-ai/dsh-skill-filesystem'
config:
providerName: riffkit-bundle
includeDefaultRoots: false
customSkillDirs: [ … the skills/ dir inside this package … ]
It inserts its own provider row rather than editing the base one, because a
patch replaces a targeted row's whole config instead of merging into it —
overriding would wipe whatever you had configured there. With
includeDefaultRoots: false this provider sees exactly one root: its own. It
cannot shadow your skills, and yours cannot shadow it.
Not using DSH?
Then you do not need this package. The skill is a single Markdown file and every harness that reads a skills root can load it directly:
mkdir -p ~/.agents/skills/riffkit
curl -sSL https://riffkit.ai/SKILL.md -o ~/.agents/skills/riffkit/SKILL.md
That path — ~/.agents/skills — is the shared root several harnesses scan, so
one copy registers Riffkit in all of them. See
riffkit/skill for the skill itself, examples,
and the install lines for Claude Code, Cursor and Codex.
How this stays current
Nothing here is written by hand. A daily workflow pulls SKILL.md,
HEARTBEAT.md and SKILL.json from riffkit.ai — the single source of truth,
served from the app on deploy — derives package.json and cordis.patch.yml
from them (scripts/generate.mjs), proves the result is still discoverable by
DSH's own provider (scripts/smoke.mjs), and publishes any version the registry
does not have yet.
The skill also carries its own 24-hour version heartbeat, so an installed copy re-syncs itself between package updates.
MIT.
原始 README: https://github.com/riffkit/dsh-plugin/blob/main/README.md ↗
同类插件
查看全部 →
deepseek-harness
从仓库或系统描述生成经过校验的自包含交互式架构图、流程图、时序图、数据流图和生命周期图。

dsh-plugin
通过 DSH MCP 客户端挂载 Ouroboros 的纯配置包,在 DSH 中提供 36 个涵盖需求访谈、Seed、执行、评估与演化流程的工具。

dsh-tongflow
基于 TongFlow 的“片场”插件,用于图片、配音、音乐与视频制作:agent 为每个资产生成 TongFlow 工作流文件(.tongflow.json)并通过 TongFlow 插件执行,内嵌工作流画布,按镜头/角色/take 组织项目,附漫剧模板;以 @tongflow 开头的会话进入 Studio 界面。

helloagents
AI 编码 CLI 的工作流层:技能、项目知识、交付检查、更安全的配置写入与可恢复执行

dsh-ai-novel-writer
安装专用 AI 小说创作预设与工作台:提供带修订号的本地项目资产、紧凑侧边工作台,以及需要原生审批的逐文件变更。

rea
用 agent 逆向任何东西:从应用行为到原生二进制