Agent 連携 · DeepSeek Harness プラグイン — 226 ページ

把某个 agent 预设下模型实际看到的系统提示词、工具目录变成可视化配置，并支持为每个子代理角色单独覆盖。不写宿主预设文件。

DSH plugin that scans other AI agents on this machine (Codex, Claude Code, cc-switch, custom sources) and one-click syncs their MCP servers and skills into DeepSeek Harness.

Choose default and per-name provider, model, and reasoning effort for newly created DeepSeek Harness Agent Teams teammates, with live team status and restart-safe route views.

多 agent 团队协作：并行调试、并行功能开发、任务协调、沟通协议、团队构成。受 wshobson/agents（38k★ MIT）启发。

多 agent 团队协作：并行调试、并行功能开发、任务协调、沟通协议、团队构成。受 wshobson/agents（38k★ MIT）启发。

DSH web plugin: a Settings → General switch that lets AgentTeams run without the Web plan-approval click.

DSH plugin: Settings → Plugins card for AgentTeams member cap and per-team concurrency cap.

Independent team coordination, dynamic routing, usage ledger, and Web panel for DeepSeek Harness

Agent terminal for the DSH web GUI: sidebar entry + center panel that runs real external agent CLIs (claude code, hermes, ...) interactively inside a local PTY, rendered with xterm.js. 100% fidelity — the page IS the agent's own TUI. Sessions survive tab switches / panel close (host-side scrollback replay on re-attach), loopback-fenced REST + WebSocket routes.

思维插件（提示词层面 MoE）：14 个思维模块按需动态加载注入 system prompt，任务相关时激活对应思维方式（感知/推演/执行/交付/反馈五层认知流）

Cross-agent local LLM token usage dashboard for the DSH web GUI: scans Claude Code / OpenCode / Cursor IDE / DSH / OpenClaw / QClaw / CodeWhale / WorkBuddy / WPS Claude / Hermes logs (plus Codex/Gemini/Cline/Pi parsers on standby), keeps a persistent ratchet ledger in ~/.agent-token-stats, and serves a sidebar analytics panel with cache-hit rate, cost estimation, period comparison, day drill-down and CSV export.

工具面分档：在 agent 作用域内按证据化的 deny 集收窄模型可见工具（lean/full 一键切换 + 审计留痕），降低工具 schema 的固定上下文成本与选择稀释。

Agent Trace — visualize an agent's reasoning, parallel tool calls, and results as an interactive DAG inside the DeepSeek Harness chat stream.

多模态视觉插件（辅助通道）：把本地图片喂给 OpenAI 兼容 VLM（默认 qwen3.8-flash）做读图/描述/双图对比——批量审图/省主会话上下文；主会话亲眼看图走官方 read_image（原生多模态）

Per-command proxy for DeepSeek Harness: the agent decides which shell commands leave through a proxy, backed by a kernel the plugin installs and supervises itself.

Give a DeepSeek Harness agent its own email inbox, with inbound mail bound to per-thread sessions.

DSH ↔ agentmemory session-memory bridge (backend-only). Mirrors session lifecycle into agentmemory with standard hookTypes (compression-friendly, dedup-safe), exposes memory_recall / memory_remember tools, and injects memory context via agent/pre-step. Configuration comes from the cordis.yml row.

Agent-plane tool plugin: connects dsh to the agentmemory REST API (search / remember / forget / health)

AgentMonitor multi-agent monitoring panel for the DeepSeek Harness — a sidebar entry that serves an embedded agent-monitor dashboard via the /agentmonitor webserver route. Ships a self-contained static dashboard (works out of the box) and lets you point dataDir at your own live data pipeline.

可扩展的 AI 群像聊天游戏框架（DeepSeek Harness 插件）：输入一个行动，多个 NPC 各自回应，带头像插图；世界、地点、NPC 与功能扩展都可在对话里让 agent 直接添加修改，无需写代码。

在 DSH Web 设置面板管理 AGENTS.md：编辑全局 AGENTS.md、编辑项目级 AGENTS.md，并按「合并全局规则」与「直接覆盖」两种方式生成项目级 AGENTS.md。仅适用于 DSH 0.2.0-rc.2。

dsh-AgentTask: 纯任务监视(胶囊 x/y + 面板三段 + 点击跳转), 全端可用

每日 AI 资讯 Agent 插件：日报(存70天)/周报/月报 · 结构化生成(合并提示词) · 真实web搜索 · host+client 双半面

AI 多角色协作开发框架：文件即通信 + 教训库 + 决策日志 + 一致性审计 + 复盘持续优化（REVIEW）。注册 ai_collab_init 工具。