dsh-prompt-firewall
面向 DeepSeek Harness 的提示词卫生、可观测性与策略中间件。
Prompt hygiene, observability, and policy middleware for DeepSeek Harness.
2926 个插件 · 第 20 页,共 122 页
按 npm 发布时间倒序
面向 DeepSeek Harness 的提示词卫生、可观测性与策略中间件。
Prompt hygiene, observability, and policy middleware for DeepSeek Harness.
用于 DeepSeek Harness 的专注、响应式 QA/聊天页面,位于 /qa。
A focused, responsive QA/chat page for DeepSeek Harness at /qa.
DeepSeek Harness 的 Sleev 路由可观测性。
Sleev routing observability for DeepSeek Harness.
服务端 DeepSeek Harness 插件,用于扫描已持久化和实时的会话日志,以查找明确的用户纠正。
Server-side DeepSeek Harness plugin that scans persisted and live session logs for explicit user corrections.
DeepSeek Harness 插件界面中视觉问题的早期保守兼容层。
An early, conservative compatibility layer for visual problems in DeepSeek Harness plugin interfaces.
为 DeepSeek Harness 保留持久化的、未发送的未来对话。
Persistent, unsent future conversations for DeepSeek Harness.
用于 DeepSeek Harness 的会话级 Chromium 运行时。
Session-scoped Chromium runtime for DeepSeek Harness.
DeepSeek Harness 插件共用的运行时辅助工具。
Shared runtime helpers for DeepSeek Harness plugins.
DSH Market 的插件端:装进 DeepSeek Harness 侧边栏的插件市场 —— 浏览 / 搜索 / 猜你喜欢 / 一键安装 / 已装管理。
DSH Market 插件端:cordis 侧边栏插件(浏览/搜索/猜你喜欢/一键安装/已装管理),核心逻辑来自 @dsh-market/core
DeepSeek Harness 的 OpenViking 记忆:持久记忆工具、技能、对话捕获与提交——并带有可完全关闭的自动 profile/recall 注入。
OpenViking memory for DeepSeek Harness: durable memory tools, skills, conversation capture and commit — with automatic profile/recall injection you can switch off entirely.
@yadsh/dsh-plugin-log 的 UI,位于两处。
UI for @yadsh/dsh-plugin-log, in two places。
DeepSeek Harness 的按会话工作区可见性范围。
Per-session workspace visibility scopes for DeepSeek Harness.
为产品的某一部分配置专属专家 agent:一个 persona、一个 scope、一个 memory namespace、一个 tool policy——在 DSH Web UI 中配置,并作为普通的 DeepSeek Harness subagent 执行。
Give a part of your product its own expert agent: one persona, one scope, one memory namespace, one tool policy — configured in the DSH Web UI and executed as an ordinary DeepSeek Harness subagent.
俄语指南:Jira 和 Confluence 的配置。
Русский гайд: настройка Jira и Confluence。
DeepSeek Harness 插件的运行时本地化覆盖。
Runtime localization overrides for DeepSeek Harness plugins.
DeepSeek Harness 代理的独立答案审核门:在候选最终答案被允许完成其回合之前,由独立审核者检查,并可将其退回修正。
Independent answer review gate for DeepSeek Harness agents: before a candidate final answer is allowed to complete its turn, an independent reviewer checks it and can send it back for correction.
dsh-kv-persist 在会话切换和重启后保持本地 LLM 会话处于热状态。
dsh-kv-persist keeps local LLM sessions warm across session switches and restarts.
面向 DeepSeek Harness 智能体的只读 git 溯源工具:模型可以针对会话仓库回答“这次改动是为哪个任务做的”——分支 → 历史 → blame → 提交——全程无需接收命令行。
Read-only git provenance tools for DeepSeek Harness agents: the model can answer "what task was this changed for" against the session repository — branch → history → blame → commit — without ever receiving a command line.
面向 DeepSeek Harness 智能体的知识库工具:模型可以查询 LightRAG 图 RAG 索引对某个主题的了解,以及该索引包含什么、健康状况如何,而无需获得 HTTP 客户端。
Knowledge-base tools for DeepSeek Harness agents: the model can ask what a LightRAG graph-RAG index knows about a topic — and what the index holds and how healthy it is — without ever receiving an HTTP client.
为 DeepSeek Harness 管理的文档流水线:agent 写入 Markdown 并收到 DOCX/PDF,交出 DOCX/PDF 并收到 Markdown,或存储在线来源的文本——wiki 附件、位于……之后的文档
Managed document pipeline for the DeepSeek Harness: the agent writes Markdown and receives DOCX/PDF, hands over a DOCX/PDF and receives Markdown, or stores the text of an online source — a wiki attachment, a document behind an…
DeepSeek Harness 插件可将体积庞大的工具输出保留在模型上下文之外,同时不会将其丢弃:大型工具结果会以 SHA-256 为键,一次性存储在本地内容寻址存储中,会话仅保留一个有界的确定性…
DeepSeek Harness plugin that keeps bulky tool output out of the model context without throwing it away: large tool results are stored once in a local content-addressed store keyed by SHA-256, the session keeps only a bounded deterministic…
@asun-labs/dsh-plugin-agent-deck 将 Codex 子任务连接到 DeepSeek Harness 对话。
@asun-labs/dsh-plugin-agent-deck connects Codex child tasks to a DeepSeek Harness conversation.
三合一控制台 · dsh-s-m-c-center。
Three-in-One Console · dsh-s-m-c-center。
根据各提供商的不同,使用 OAuth、设备码或 API key 将 ChatGPT、Claude、SuperGrok、OpenCode Zen/Go、OpenRouter、Nous Portal 和 Infron 作为模型提供商。
Use ChatGPT, Claude, SuperGrok, OpenCode Zen/Go, OpenRouter, Nous Portal, and Infron as model providers with OAuth, device code, or an API key according to the provider.