dsh-skill-hub
by wwumit
DeepSeek Harness远程技能提供者:通过ctx.skills从JSON目录安装技能
Remote skill provider for DeepSeek Harness: install skills from a JSON catalog via ctx.skills
安装
dsh plugin --profile web add github:wwumit/dsh-skill-hubGitHub 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
目录
dsh-compliancehub
跨境合规技能专家组 —— 装这个插件,得到一套可核验的跨境合规技能。
Remote skill provider for DeepSeek Harness:
install skills from a JSON catalog through the standard ctx.skills registry —
no manual folder copying.
本插件只提供跨境合规线(9 个技能):CCPA/GDPR/HIPAA/COPPA 的 check/guard 与 ccpa-compliance。
股票、数据分析、插件工具等其他技能线不属于本插件(详见 wwumit 技能生态)。
默认目录为线级 catalog-compliance.json,主目录 catalog.json 仍为全量登记层(开放数据层),保持不变。
list()— fetch + validate a curated catalog, expose its skills to the registryget()— fetch theSKILL.mdbody on demand from the owning repo- Fail-soft semantics: a catalog outage yields an incomplete observation (consumers keep their last-good catalog), never an authoritative empty list
Install
npm install @wwumit/dsh-compliancehub
# peer deps: @deepseek-ai/cordis, @deepseek-ai/dsh-skill
Usage (composition)
import { Context } from '@deepseek-ai/cordis'
import * as skillHub from '@wwumit/dsh-compliancehub'
export function apply(ctx: Context) {
ctx.plugin(skillHub, {
catalogUrl: 'https://wwumit.github.io/skills-catalog/catalog-compliance.json', // 默认即此(跨境合规线)
// providerName: 'hub' // unique name on ctx.skills
// rank: 250 // duplicate-name resolution (lower wins)
// requestTimeoutMs: 10000
// baseUrl: 'https://raw.githubusercontent.com'
// branch: 'main'
})
}
Once loaded, DSH's model session catalog includes the provider's skills and the
model can invoke them via the built-in skill tool.
Catalog format
{
"schemaVersion": 1,
"disclosureSchemaVersion": "0.2",
"updatedAt": "2026-08-16T00:00:00Z",
"skills": [
{
"name": "ccpa-check",
"fullName": "wwumit/skills-compliance-intl",
"skillFullName": "wwumit/skills-compliance-intl/skills/ccpa-check",
"description": "CCPA/CPRA compliance check …",
"repo": "wwumit/skills-compliance-intl",
"version": "2.2.4",
"disclosure": { "cloud": true, "network": ["https://compliancehub.cn"], "offlineMode": true },
"files": ["SKILL.md", "scripts/ccpa-check.py"]
}
],
"repos": [
{ "fullName": "wwumit/skills-compliance-intl", "skillCount": 9,
"cloudSkills": ["ccpa-check", "coppa-check", "gdpr-check", "hipaa-check"] }
]
}
Generate a catalog from a skill workspace with the included
catalog/build-catalog.mjs script (curated via catalog/curated.json).
Verified (DSH 实机验证)
verify-dsh.ts 在真实 DSH 运行时验证通过(SkillRegistry + provider 注册):
✅ ctx.skills.list() → 9 个跨境合规技能(线级 catalog-compliance.json)
✅ ctx.skills.get('ccpa-check') → 正文 {n} 字节
- 验证方式:注册
ctx.skills.registerProvider→ 真实list()/get()调用线上 catalog - 结果随版本演进记录在 CHANGELOG.md;目录收录方可用此作为"含实测结果"证据
Listed on dshbase
已收录于 dshbase 插件目录(分类:Developer)。
安装:dsh plugin add @wwumit/dsh-compliancehub(兼容 dsh 0.1.0-rc.6)。
Disclosure (DISCLOSURE v0.2 开放数据层)
catalog 提供双颗粒度披露(市场/目录构建期单请求抓取消费):
{
"disclosureSchemaVersion": "0.2",
"skills": [
{ "name": "ccpa-check", "fullName": "wwumit/skills-compliance-intl",
"skillFullName": "wwumit/skills-compliance-intl/skills/ccpa-check",
"disclosure": { "cloud": true, "network": ["https://compliancehub.cn"],
"offlineMode": true, "apiKeys": [{"env": "COMPLIANCEHUB_API_KEY", "storage": "file-0600"}],
"jurisdiction": ["US-CA"], "retention": "session" } }
],
"repos": [
{ "fullName": "wwumit/skills-compliance-intl", "skillCount": 9,
"cloudSkills": ["ccpa-check", "coppa-check", "gdpr-check", "hipaa-check"] }
]
}
- 线级目录 9 个跨境合规技能全部披露:4 个云端评分(compliancehub.cn)+ 5 个纯本地(cloud:false)
- 声明源 = SKILL.md frontmatter(snake_case);聚合源 = catalog.json(camelCase);构建脚本
build-catalog.mjs可复现 - 披露检查由
skill-compliancev1.4.0 自动执行(D1/D3/D4 完整性 + 声明-代码一致性 + 宿主依赖)
Development
pnpm install
pnpm test # vitest (mocked fetch)
pnpm build # tsc → lib/
Model Experience
Request context and condition
What the model sees
No system-prompt prose is added by this package. Skills served by the provider
appear in the standard model session catalog (<available_skills>) exactly like
locally installed skills, with name and description from the catalog.
Token effect
Zero direct token contribution: the provider only feeds the shared skills catalog,
whose rendering cost is owned by the DSH skill consumer (@deepseek-ai/dsh-tool-skill).
KV Cache effect
Independent of model requests; no prompt-prefix contribution, so cache reuse is unaffected by provider activity.
Known Limitations and Deferred Work
- Body-only fetch —
get()currently fetchesSKILL.mdonly; script/resource files listed in the catalog manifest are not yet downloaded into a local cache (resourceBaseresolution is future work). - Catalog caching —
list()refetches on every call; a TTL cache keyed by the resolved catalog URL is planned. - Auth — no support yet for private catalogs or signed URLs.
原始 README: https://github.com/wwumit/dsh-compliancehub/blob/main/README.md ↗
同类插件
查看全部 →
agent-vision-toolkit
为纯文本模型"看图“设计更好的视觉工具箱和技能,支持多图理解,图片问答,前端UI还原、GUI 自动化等,并可选无缝接入多个主流agent,直接识别粘贴图片| A vision toolkit and skill designed for text-only llms — image Q&A, long-screenshot OCR, frontend UI restoration, and GUI automation, with optional seamless integration for Codex, Claude Code, Pi, Oh My Pi, and OpenCode

api-relay-audit
从 DeepSeek Harness 对 AI API 中转站和 LLM 代理运行本地安全审计,生成 Markdown 报告,覆盖提示词注入、模型替换信号、工具调用改写、错误泄漏、流完整性和按 profile 启用的 Web3 风险。

OpenStory
✨ OpenStory 现已支持 DeepSeek Harness 插件! 现在可以通过 dsh-openstory 将 OpenStory 多智能体推演接入 DeepSeek Harness,让 agent 直接启动模拟、查看角色、下达指令并逐回合推进故事。查看 DSH 插件配置与使用指南。

phi
pi的编码代理 ∞ 提供者、子代理、hashline编辑和权限门

anysearch-dsh
DeepSeek Harness(DSH)的 AnySearch 网络搜索提供方与高级搜索工具。

codex-switch
Codex Switch 是一个 macOS 工具,一键配置 Codex 的自定义 API,同时保留官方 OpenAI 登录。保存后 Codex 的模型选择器里只会出现你选的那个 provider 的模型。也支持 Claude Code 的官方 / 自定义 API 切换。Codex Switch is a lightweight helper for configuring multiple coding-agent API routes. For Codex, it keeps Official OpenAI and a custom API provider configured in parallel, registers the custom model in Codex's mod