플러그인

DeepSeek Harness용 GPU 인식 실행 레이어: gpu_status / gpu_exec / gpu_run_bg 도구, 단계별 GPU 컨텍스트 주입, CUDA_VISIBLE_DEVICES 카드 자동 선택

DeepSeek Harness 플러그인: OpenAI 호환 엔드포인트의 Qwen3 계열 모델을 위한 요청별 사고 모드 전환(enable_thinking)

DeepSeek Harness용 DeepSeek 사용량 모니터: 모델이 호출 가능한 도구로 제공되는 계정 잔액 조회와 토큰 사용량 요약

DeepSeek Harness 이중 기능 플러그인: Web 설정의 시각적 플러그인 관리자 탭 — 모든 플러그인 행을 켜고 끄고, 설치된 플러그인(공식 @deepseek-ai/* 항목 포함)을 npm 레지스트리와 대조해 확인하고, 업데이트하고, 원본 파일을 복구하며, 하네스 자체도 업데이트

MCP를 통한 호스팅 도구 서버: 인보이스와 은행 명세서 추출(코드 내 산술 검증 포함), 정수 센트 장부 대사, PDF와 표 변환, SQL 방언 변환, 중국 이동 네트워크 관점에서의 중국 본토 도달성 프로브.

코딩 에이전트를 위한 디자인 I/O: 선언(spec/domain/craft/design/components/template)과 계약(skill/evaluator)을 통한 제어 가능한 UI 생성. 콘솔, 대시보드, 에이전트 운영, CJK 우선 앱 등 제품 UI에 활용

/repro는 최소한의, 비밀 정보가 제거된, 재현 가능한 문제 번들을 내보냄: 세션 로그, 실패한 명령, Git diff

DeepSeek Harness용 재사용 가능한 Markdown 프롬프트 프로필, 턴별 모델 선택, 인자 치환, 상태 복원 지원

텍스트 위생 관리를 dsh 플러그인으로 제공: 신뢰할 수 없는 텍스트 정제, 보이지 않는 문자 검사, LLM 포맷팅 정리, CSV 수식 인젝션 이스케이프

DeepSeek Harness의 모든 슬래시 명령어를 사용 빈도순으로 정렬해 실행 가능한 버튼으로 나열하는 플로팅 명령 데크

Fusion for DSH in one bundle: the upstream MCP tools with their exact schemas and an editable endpoint, plus the fusion-harness CLI bridged as tools.

Local dataset analysis tools for DSH — ingest once, keep large tables on disk, call small JSON tools

ROFilter reverse-osmosis design MCP server (rofilter.cn): membrane library, water analysis & full system design. stdio MCP (Coze/Doubao, Qwen, WorkBuddy, Claude, Cursor), Streamable HTTP, DeepSeek Harness bundle, plus Agent Skill zips.

Reusable report automation skill with target-template and AI-generated standard-report flows

Adds a right-sidebar tab to the DSH web UI with the daily Japanese anime broadcast calendar from Bangumi (api.bgm.tv), grouped Monday through Sunday with today highlighted; hovering a card shows a details popover (cover, score, rank, air date, episode count, tags and synopsis), and clicking opens the title on Bangumi; the host caches the calendar for 10 minutes and subjects for 24 hours on disk so the last successful table still displays offline, and falls back to common local proxy ports when the direct connection fails.

A desktop pet for DSH: a blue whale girl with 104 animations and built-in voice. It is standalone, but its pet runtime is a rework of dsh-dafeiyu — disable the upstream dsh-dafeiyu before installing, or both will run and fight over the control port. Windows x64 only, and the first launch downloads about 580 MB of assets.

Simulated FX trading arena played against an algorithmic whale-chan opponent: a live canvas candlestick chart over five instruments (EUR/TOKEN, GBP/TOKEN, TOKEN/JPY, gold, and Whale/TOKEN, her own coin she never shorts), selectable leverage from 10x to 500x, and two modes — a timed mode with 90s to 500s rounds settled on equity, and an untimed goal mode that starts with $300,000 and ends when someone first reaches $20M or goes into debt (equity below zero, no maintenance margin). The match runs in a module-level engine, so switching to a conversation does not stop it or restart it. Margin and spread are modeled, prices are a local random walk with occasional news shocks, and nothing touches the network or real money.

SSH-backed remote directories and linked local⇄remote workspaces: read-only remote browsing, an editable linked workbench, bidirectional SFTP transfer, mirroring and a linked_workspace tool. Sidebar pages require the dsh-better-sidebar plugin (v0.24+); without it only the Settings pages load.

Manage agent-team roles in the dsh Web GUI: each role becomes its own selectable agent preset with per-role tool, skill and MCP switches.

A five-stage engineering loop for agent-built projects — spec, build, verify, ship, learn — with one artifact and one gate per phase, so an agent always knows the single next action. Role policy, project templates and deployment sit behind explicit ports: swap the implementation without rewriting the loop. Its scaffold generator lays out the internal layers of a project — `init --layout {none,webapp,service,plugin} --stack {python,node,none}` — from nine layers sampled out of three unrelated production projects: entry docs, an OpenSpec plan face, a spec truth layer (INDEX, one file per domain, TESTING, CONSTITUTION), a wiki split into knowledge / progress / runbook, the six `.agents/` slots, server+web, scripts/deploy/config, tests+e2e and a gitignored out/. Non-none layouts also gate the directory contract and compile every source root. Ships a Python CLI (init/next/check/ship/new) plus a local command adapter and an SSH/systemd adapter with release directories, an atomic current/previous switch, a deploy lock, a commit-versioned JSON probe and automatic rollback on a failed probe; bootstrap and teardown provision and remove the unit and the reverse-proxy block idempotently and refuse to delete a root they did not create. Includes a runnable example workspace and a skill that keeps the agent on one phase at a time. Requires Python 3.10+ and PyYAML; remote deploys need SSH access, rsync, systemd and Caddy on the target.

Editable PowerPoint authoring workflow (a DSH-skill adaptation of icip-cas/PPTAgent) with concise Chinese business styling, HTML slide rendering and PPTX export review. Linux/WSL or macOS only; first use requires LibreOffice, Chinese fonts and a one-time setup step that installs Python/uv and Playwright Chromium.

Agent application evaluation toolkit for DeepSeek Harness: offline/online suites judged from audit side effects, baseline diff with zero safety-regression tolerance, and mutation tests proving the evaluators themselves catch regressions. Framework-agnostic - needs a callable chat endpoint and queryable tool-call records.

Left-sidebar PPT workbench: imports an existing PPTX element by element (theme colours, gradients with alpha, rounded shapes, master/layout decorations, chart pages rasterized) and renders it as real 16:9 slides, lets you edit text and tables directly on the page, and hosts an AI chat column that rewrites the same deck.json.

Text-to-speech service with Doubao TTS 2.0, Alibaba Bailian and OpenAI-compatible providers and a user-chosen voice; the agent speaks through tts_service_say, split into segments with an optional tone each, and other plugins can reuse ctx.ttsService.