Plugins

Failure-recovery memory: search and record failure-recovery lessons from real engineering sessions, with BM25 + semantic RAG retrieval and a lessons knowledge base.

Persistent multi-model squads configured in Settings and selected in the Composer, with per-member model/tool policies; the main Agent dynamically plans each send as a bounded DAG with bounded reviewer/repair loops, while Run Center tracks retries and per-member tokens from official Harness usage events.

Pop-out sidebar: lists artifacts and a file tree with multiple preview formats, pops out into a separate browser tab you can drag to another monitor, and coexists with other sidebar plugins simultaneously.

DSH↔Godot engine runtime bridge: launch and drive a running Godot 4.x game through its in-game TCP interaction server — 16 tools (runtime control via ~130 game commands, scene/UI inspection, GDScript eval, input simulation, screenshots, headless static ops, project editing, export, script validation, engine-path setup), replaces godot-mcp.

Restart recovery for DeepSeek Harness web: after a crash or restart, an interrupted agent turn continues automatically (host-side agent/created listener, no browser timing races).

Pops up a mini-game menu (wordle, match-3, extensible) while the model generates.

Expert-mode agent preset for DeepSeek Harness (v0.3.0, bilingual EN/ZH): a chief coordinator plus 11 domain-expert subagents with automatic task delegation. Features: Five-Anchor constraint (review/convergence/anti-drift/collaboration-check/resource-awareness per turn), Near-distance Guidance (identity/task/output template per expert), expert persistence, cross-expert communication protocol, cross review, and experience pool. Experts: data analyst, copywriter, legal review, product manager, frontend, UI/UX, architect, social media ops, growth hacker, quant finance, finance.

SSH ops terminal for DSH Web UI: drive a connected server from the main conversation (ssh_connect/ssh_exec/ssh_read/ssh_write/ssh_disconnect) while keeping an interactive xterm.js terminal on the right.

Recurring loops: `/loop` command + loop tool + activity status bar.

Docker tools for agents: ps/images/logs/inspect/exec/manage with JSON output, env-configurable CLI path, approval-gated exec and zero runtime dependencies.

Evidence-first operating kit: six read-only tools (capability catalog, staged workflow plans, packaged skill reader, bounded local memory search, repository release audit, release checklist) and five packaged skills - plans and audits, never remote writes.

One-click health check for the DSH Web GUI: a sidebar button runs the ops kit's check-health.ps1 through a plain HTTP route (independent of the tool registry) and renders a structured report card; colors follow the active theme, Windows/PowerShell.

Windows desktop shortcut for dsh web: launches the server, auto-opens the browser when ready, and stops the service when the Harness window closes.

Detects DSH plugin conflicts (duplicate mounts, failed/pending plugins, duplicate config row ids) at startup, auto-stops or restores the affected plugins, and shows a conflict report popup in the web UI. DSH 启动时检测插件冲突并自动停用/恢复，弹窗汇报冲突详情。

Mention a server with @ in a DeepSeek Harness session and run commands on it over SSH, behind a two-tier approval gate that fails closed.

A project-agnostic multi-agent collaboration protocol for the DeepSeek Harness: loop guards that stop runaway agent loops, a six-field handoff format, risk-tiered review routing, and a fixed verify→commit→push→review→close sequence. One skill instead of the same rules copy-pasted into every agent's instructions.

SSH remote ops for DSH: pair servers over SSH or a one-time-code remote-hostd, then run commands, manage background jobs, and read or write remote files with reviewable changes (host_pair/host_bash/host_jobs/host_read_file/host_write_file/host_review_changes).

DeepSeek Harness (dsh) Windows notification plugin: a native toast when the agent stops running (finished / aborted / error / waiting for your choice / session closed), plus a system-tray icon. Pure host-side, zero dependencies, zero build.

Design I/O for coding agents: controllable UI generation via declarations (spec/domain/craft/design/components/template) and contracts (skill/evaluator). Use for product UI—console, dashboard, agent-ops, CJK-first apps.

Typed, guarded container control: ps/logs/inspect/exec/start/stop and compose up/down with JSON output, project-aware targeting, and approval-gated destructive ops.

Enable/disable MCP servers directly from DSH Settings: a 'MCP Servers' page that stops/starts each @deepseek-ai/dsh-mcp-client connection live and persists the change across restarts.

Enable/disable DSH plugins directly from Settings → Plugins: an additive 'Enable/Disable' tab that stops/starts loader entries live and persists the change across restarts.

Auto-continue an interrupted dsh session after a web restart by injecting a "continue" follow-up; one-shot per process, no duplicate-injection loops.

DSH Web GUI paperclip button: a single 📎 button in the composer that opens a file picker and drops file content into the textarea.