dsh-rayrun-computer: Create a Cloud Linux Desktop in the DSH Side Panel and Let the Agent Work on the Same Machine
dsh-rayrun-computer is a DSH client plugin maintained by RayrunOS. It creates and manages Rayrun cloud Linux environments in the sidebar, embeds a noVNC desktop, and provides tools such as rayrun_sandbox, rayrun_exec, and rayrun_workspace. It supports session-based cloud/local routing, token storage via Harness vault, and local-first push/pull verification synchronization. It must be installed from the source directory under the DSH 0.2.1-rc.5 environment with a Rayrun account configured. There is no outbound network by default, destroy is irreversible, and the bundle code runs within the DSH process.
Read Moredsh-qoder-cloud-agents: Running DSH Sessions and Delegation Tasks in Qoder Cloud Sandboxes
dsh-qoder-cloud-agents is a DeepSeek Harness workflow plugin maintained by QoderAI. It runs full cloud-mode sessions within the Qoder Cloud Agents sandbox, delegates subagent_qoder_cloud_agents subtasks, or manages the environment and Agents through /qoder-cloud-agents. When the workspace is a git repository, it mounts the origin remote and pushes changes to the qoder-cloud-agents branch. Based on the official README, this article covers prerequisites, installation via npm pack, loading the tgz into the profile, ~/.dsh/.env tokens and cordis.patch.yml configuration, as well as typical usage, costs, and security considerations for the three entry points.
Read MorePitfall Avoidance Guide for Home Renovation: Automating the Screening and De-risking Process with Agents
The Home Renovation Company Pitfall-Avoidance Guide (handle: @user_cf67e960/zhuangxiu-gongsi-bilei; v1.0.4; source: community; category: life-service; API key not required) is a community Skill designed for homeowners selecting renovation/renovation firms. It turns the five stages—choosing a company, risk screening, reviewing quotations, auditing contracts, and handling mid-project price increases—into a reusable checklist framework. The official description states that it embeds 181 core reference documents. Usage is straightforward: ask questions in natural language, such as how to shortlist companies, which companies to avoid, how to identify missing line items in a quote, which contract clauses must be explicitly stated, and how to deal with on-site cost escalations during construction. This article summarizes the verified metadata and coordinates, notes the limitation that no installation command is provided (the official page does not offer a verifiable installation method, so no command is listed; it is only described as being loaded according to a standard SKILL.md directory layout), and outlines the target audience and constraints: the Skill provides methodology and does not replace on-site construction inspections and credential verification; the number of references is per official claims and was not independently verified on a per-item basis. Captured page data shows a score of 100000, 1,338 downloads, and 0 installs.
Read Moredsh-bili-eyes: Enabling DSH Models to "See" Bilibili Videos
dsh-bili-eyes is a DSH networking plugin maintained by QiuYeDx that only recognizes Bilibili links. Through tools such as bili_look, it returns storyboard images along with titles, popular comments, and trending danmaku keywords. Optional login enables AI-generated subtitles and summaries. The plugin is read-only by default, and ffmpeg can be installed with one click on the settings page. The installation command is `dsh plugin --profile desktop add dsh-bili-eyes`, requiring DSH version 0.2.1 or later. Write operations, inbox access, comment moderation, and similar features must be enabled according to the permission table on the "Bilibili Partner" settings page. It is recommended to use a secondary account and to review the source code and potential risks yourself.
Read MoreEngineering Design Solutions: Ensuring Bids and Proposals Withstand Evaluation Scrutiny
Introduce the community Skill "Engineered Solution Design" (@user_ca0a159e/design-engineered-solution, v1.0.0, business-ops, rating 100000/downloads 1303). It employs engineering methodologies to draft, refine, or review solutions, bidding technical and engineering proposals, covering sections such as Introduction, Overview, and Implementation Pathway. The objective is to ensure that each step is acceptable, risks are mitigated, commitments are verifiable, and the entire process is traceable. Core capabilities include drafting and refining proposals, identifying shallow or unstructured content in reviews, rewriting proposals from a client perspective, and completing acceptance and risk mechanisms. It does not require an API Key. The text provides the official address and applicable scenarios, and notes limitations such as the lack of an installation command and the requirement for manual review of business data.
Read MoreThu-tok-auto: Automatically Obtaining and Renewing Tsinghua MadModel Tokens in DSH
thu-tok-auto is a DeepSeek Harness profile bundle plugin maintained by OverDustD7, designed for madmodel.cs.tsinghua.edu.cn. It facilitates tiered JWT renewal via Get/Auto modes, writes to MADMODEL_API_KEY, and syncs with llm-pi-ai.providers.madmodel, supporting SSO replay and a unified authentication login fallback for manual Get requests. Based on the official README, this article covers key usage points including installation commands, renewal strategies, UI status semantics, credential YAML configuration, permissions, and known limitations.
Read MoreNeko_02 (dsh-usage-overview): View Account Balance and Usage Heatmap within DSH Desktop
Neko_02 (dsh-usage-overview) is a DSH desktop community client plugin maintained by nTillMeetu. It replaces the "Settings → Account & Balance → Usage Query" section from an embedded platform page with an in-app panel, displaying balance, cumulative consumption/tokens/requests, monthly calendar heatmaps, and daily breakdowns. It supports API Key filtering and current month CSV export. The Host side calls the Platform usage API to aggregate data, with tokens remaining on the Host. Installation requires cloning the repository and running `node install.mjs`, followed by a DSH restart. Licensed under MIT 1.0.1, this is an unofficial product; changes in the DSH API may lead to compatibility issues.
Read Moredsh-freecad: Read/Write STEP and Generate First Angle Projection Drawings in Local FreeCAD
dsh-freecad is a DSH client bundle maintained by Negulons: geometry processing is performed within the FreeCAD process, providing tools such as freecad_inspect_step, freecad_edit_step, and freecad_step_to_drawing to read STEP files and generate SVG/DXF/PDF/PNG machining drawings. It supports installation via a plugin page or npm and requires FreeCAD 1.0/1.1 along with freecadcmd. This is not production-grade drafting, and data remains local on the machine.
Read MoreDSH-GitHub-Trending: Generating and Auto-PRing Daily Chinese GitHub Trending Reports in DSH
The DSH networking plugin maintained by morehao, @morehao/dsh-github-trending, schedules the local trendradar core to fetch GitHub Trending on a daily/weekly/monthly basis, generates Chinese introductions and summaries, optionally subscribes to Release summaries, automatically commits and opens PRs. The plugin handles configuration, scheduling, and the workbench (overview, rankings, run records, settings, subscriptions) without modifying the core output. Installation: dsh plugin --profile <profile> add @morehao/dsh-github-trending; requires DSH 0.2.1-rc.3+, Node ^22.20 or ≥24, git ≥2.30.
Read MoreLinkPix: Introduction to the AI Viral Video Replication and Audio Extraction Skill
**Introduction** Teams engaged in short video marketing often face two recurring challenges: upon seeing a viral video, they wish to replicate its pacing and presentation style, but manually analyzing shot breakdowns, beat synchronization, and transitions consumes significant time. Furthermore, after acquiring source material, they need to extract audio separately for re-editing or multi-channel distribution. The LinkPix Skill introduced below integrates these two processes into a single tool. **What is this?** LinkPix (Skill Full Name: AI Viral Video Replication & Audio Extraction | LinkPix) is an Agent Skill designed for short video creators and e-commerce sellers. Identified as `@qinghu-ai/linkpix-viral-video-toolkit`, it is currently at version 0.1.8 and is published and maintained by Qinghu AI.
Read MoreStock Industry Chain Bottleneck Hunter (White Hair Investment Research Method): Turning Investment Research Agents into Supply Chain Bottleneck Hunters
Released by SkillHub community user @user_ac35695d, this Skill is version 1.0.0, categorized under "professional," and marked as community-sourced. It is designed to transform investment research Agents into supply chain bottleneck hunters, covering traceable source research, market and thematic scanning, AI/semiconductor/tech value chain mapping, A-share/HK stock/US stock screening, thesis stress testing, and Serenity-style research dialogue. Triggered by natural language commands such as "View like Bai Mao," "Deep Research," "Industry Chain/Blockage Point/Bottleneck," "Rank Candidates," and "Challenge This Thesis," it outputs plain-language reasoning, prioritized research, evidence chains with risks, and next-step verification steps. Officially, it explicitly supports research only and does not execute trades. Please note: The specific installation commands and configuration fields were not provided on the official page verified this time, so confirmation from the publisher is required before installation. Output quality depends on the Agent's upstream retrieval and data source coverage; bottleneck assessments are research hypotheses rather than conclusions. Ratings and download counts are subject to the real-time figures on the official page.
Read Moreds-pstack: Integrating pstack Workflows into DeepSeek Harness with Sub-agent Model Routing from Your Selected Pool
ds-pstack, maintained by monkeyin92, ports Lauren Tan's 58 pstack skills to DeepSeek Harness, retaining playbooks such as /poteto-mode and /interrogate. Unlike the Cursor version, the sub-agent model slug is not hardcoded; instead, a deterministic router selects from a model pool configured in Settings based on eight task intents, and the classification step does not invoke a model. Installation is performed with `plugin_manager install_bundle github:monkeyin92/ds-pstack#main`; it requires enabling Subagent model selection and running `/setup-pstack` to choose a budget. The bundle includes two entries, ds-pstack-skills and ds-pstack-router, and is MIT licensed; the current version is 0.1.0.
Read MoreNews Aggregator: Summarize Domestic and International Technology and Military News into a Key Points Brief
News Aggregator (@httencn/news-aggregator, v1.0.3, MIT-0) codifies the aggregation of domestic and international technology and military news into a SKILL.md workflow. It enforces a fixed four-step process: Search, Filter, Organize, and Output. The news sources are hardcoded into four distinct groups: Domestic Tech, Domestic Military, International Tech, and International Military. Its credibility rules differentiate between official reports and forum reposts, generating a fixed-format Markdown bulletin brief. The documentation outlines the specific problems it solves, built-in sources, output formatting, installation via `openclaw skills install`, and typical usage, while noting limitations such as non-configurable sources, lack of scheduling, and absence of built-in translation.
Read MoreDeepSeek Cost Meter: Per-round Model Cost Estimation in Harness
DeepSeek Cost Meter (deepseek-cost-meter) is a community DSH plugin maintained by luomuwen under the MIT license. It displays estimated model costs based on API usage and configured pricing below each response round and in the input box footer, supporting DeepSeek's peak/off-peak pricing as well as multiple channels including Kimi and MiniMax. Install with the command: dsh plugin --profile desktop add deepseek-cost-meter. It requires no API key and makes no additional model requests. Custom pricing can be added via additionalRates. The estimation excludes costs for tools and sub-agents, and may differ from account billing statements.
Read MoreDesktop Computer Use: Semantic Desktop Control Skill for Windows
Desktop Computer Use is a community-sourced Agent Skill (Coordinates @user_bf9357fd/desktop-computer-use, version 1.0.4, category ai-agent) focused on semantic desktop control for Windows: enabling AI to see and operate desktop applications and browsers as a human would, while supporting both native GUI and browser CDP channels. It is tagged requires_api_key=false. Starting from developers’ pain points, the article outlines its core capabilities, installation and activation methods (the official page is unreachable, and the installation command and configuration fields have not been verified; only the official URL is provided), as well as typical usage and applicable scenarios. Note: It is community-sourced, has zero installs, and is limited to Windows; validate independently before adopting.
Read Moredsh-mood: Adds Six-Dimensional Moods and Cyber Diaries to DeepSeek Harness
dsh-mood (Yuyu Mood Engine) provides DeepSeek Harness with a six-dimensional event-driven state machine, a 22-word emotion matrix, per-turn tone injection, optional sampling temperature intervention, scheduled diary entries, and a right-column mood panel. Based on the official README, this article explains ZIP installation, step-by-step configuration of cordis.patch, dsh plugin install, and full restart procedures, while summarizing the interaction limits, local HTTP interface, and simulated disclaimer, enabling a fun dress-up style experience without compromising security or factual accuracy.
Read Moredsh-mcp-status: Check MCP Server "Available Now" Status in Chat Header
When DSH starts, it connects to MCP. When services are abnormal, tools silently disappear or report transmission errors, and configuration files cannot reflect real-time status. dsh-mcp-status is a MIT client plugin by lobichshev-ops that provides an MCP button in the session top bar: it detects each server through initialize and tools/list handshakes (not port checking), displays status, response time, and tool count, and caches the results. Installation is done via pnpm link in the profile and bundle list registration, with optional configurations such as intervalMs. It is suitable for multi-MCP and EDT scenarios; the plugin runs with DSH process permissions, so please review the source code before use.
Read Moredsh-codebuddy-link: Invoking Tencent Models in DSH via CodeBuddy CLI
`dsh-codebuddy-link` is a DSH Cordis client plugin maintained by liubaicai. It integrates the Tencent CodeBuddy CLI into LlmAdapter via ACP stdio, supporting multi-account management and independent delegation of the `codebuddy_ask` function and the `/codebuddy` slash command. Installation requires Node.js 22.18+ and a globally installed CodeBuddy CLI; the recommended installation method is `dsh plugin add git+https` (pinned to v0.3.30). The settings panel manages models and candidate fallbacks, while credentials are stored in `DSH_HOME`. Note that ACP and permission bypass override sandbox restrictions, making it suitable for local, trusted environments only.
Read Moredsh-zh-layer: DSH Switchable Bilingual Overlay
dsh-zh-layer is a DSH client plugin maintained by Lin-AxiomSight. Through runtime metadata wrapping and front-end rendering, it provides Chinese names and descriptions for plugins, slash commands, and Skills. It supports three display modes — zh / both / off — without modifying the DSH source code. AI translation must be triggered manually, defaults to deepseek-flash, and submits only untranslated entries. Configuration and dictionary files are stored at ~/.dsh/zh-layer-config/. Based on the official README, this article covers installation commands, typical usage, privacy and cost considerations, and known limitations.
Read MoreApple Design Style: Implementing Minimalist Human-Centered Technology from Website Aesthetics to Web Pages
Codify the Apple website's color schemes, typography, layout, interactions, imagery, and copywriting rules into a Skill, featuring a built-in page generator and a style checker. v1.2 adds 3 example pages, enhanced templates, and 38 tests.
Read Moredsh-corner-gauge: Floating Balance Panel for DSH Corners with Interchangeable Skins
dsh-corner-gauge is a DSH Web client plugin maintained by kkppshuai-svg that displays a draggable balance badge in the corner of the interface. The `skin` component manages the border, while `elements` manages the content. By default, it fetches the DSH account balance; on failure, it displays `--` and retries. It supports the mclaren/minimal skins, local bridge-based reading of usage from Codex/Claude/Doubao, and a configurable sports car easter egg. Installation requires copying to the profile's node_modules, registering bundles in package.json, and restarting the App (rather than using `dsh plugin add`). Compatible with DSH >=0.2.1-rc.2 <0.3 and Node >=18. MIT licensed.
Read MoreExpert in Document Polishing and Layout: De-AI-ify and Refine Chinese and English Texts — Skill
This article introduces the SkillHub community Skill "Document Polishing and Formatting Expert", with coordinates at @user_e9988879/document-polisher, version 1.0.0, and categorized under content-creation. Designed for intelligent polishing of Chinese and English documents, its capabilities include removing AI-like tone, grammar correction, style standardization, layout optimization, and conciseness enhancement, making it suitable for workplace writing, academic papers, and self-media content creation. The article has verified the official address, source attribution, and page metrics, and notes that the provided documentation does not include a directly copyable installation command. Actual activation depends on the Skills loading rules of the host environment, such as Cursor, Codex CLI, or Claude Code.
Read Moredsh-computer-use: Enabling DSH Agents to See and Operate Real Desktops on Windows
dsh-computer-use is a DSH workflow plugin maintained by Iroul1212 (v0.2.0, MIT licensed). It registers only five tool categories—computer, app_window, ui_element, clipboard, and app_skill—on Windows 10/11, driving the real desktop through a persistent PowerShell process. Screenshots are routed through an attachment pipeline with pixel budget control; it supports UIA text and 8 built-in app skills. The overlay and Esc emergency stop work in conjunction with dryRun, approval, and blocklist mechanisms. Installation is available via npx dsh plugin add using file, github, or tgz paths; a DSH restart is required after adding the package.
Read MoreChinese Communication Poster Design Expert: Ensuring Rigorous Pre-Publication Visual Acceptance
This article introduces the community skill "Chinese Communication Poster Design Expert" on SkillHub (@user_cae83fba/poster-design-pro, version 2.0.0, category: design-media, tag: requires_api_key=false). This skill is designed for pre-publication visual QA and delivery review of Chinese communication posters, covering four scenarios: WeChat Official Account covers, social media covers, and event, exhibition, and product launch posters. Its core capabilities consist of four acceptance gates: four-tier information architecture, thumbnail readability verification, QR code and CTA gating, and copyright and compliance review. The skill explicitly performs only evaluation and acceptance; for image generation, it relies on tool routing rather than building its own. Therefore, it is not applicable for general image generation, complete brand VI systems, replicating official brand styles, or imitating unauthorized materials. The text explains the activation method for this type of skill based on the general SKILL.md format, and notes that official installation commands and examples were not obtained during this scrape, meaning the relevant commands must refer to the actual page.
Read Moredsh-ieee-weekly: IEEE Xplore Early Access Keyword Weekly Report
dsh-ieee-weekly is a plugin for the DSH networking tool, maintained by hxxhxhxx. It discovers new DOIs from IEEE journals through Crossref via a windowed approach, fetches abstracts and Early Access flags from Xplore, and then analyzes each entry using keywords and DeepSeek. It generates a Markdown weekly report and optionally sends it via SMTP, with DOI deduplication to prevent duplicate pushes. The npm package name is dsh-ieee-weekly; after installation from the marketplace, it can invoke the ieee_weekly tool. It also supports the ieee-weekly CLI. Configuration is centralized in config.json, with the data directory defaulting to ~/.ieee-weekly/. It is MIT licensed and requires Node.js ≥ 20.
Read More