DSH-oopz-voice: Enabling DSH Agents to Enter Oopz Voice Rooms and Speak
dsh-oopz-voice is a DSH workflow plugin maintained by Kingyibin. Using local SenseVoice ASR, an OpenAI-compatible LLM, VITS TTS, and Agora streaming, it enables an agent to listen, think, and speak in authorized Oopz voice channels, and supports yielding and catch-up listening. Install it with `dsh plugin --profile desktop add dsh-oopz-voice`; requirements include Node 20+, Python 3.10+, real Chrome, and an Oopz account. Perceived latency is about 1.6 seconds. It connects via reverse-engineered access to the private protocol rather than through an official API, so please read the compliance and privacy notice before use.
Read More1688 Best Seller Selection | Guo Yiyi: Delegating Sourcing Decisions to Structured Data
1688 Best-Selling Product Selection | Guo Yiyi (v1.0.1) is a community Skill designed for 1688 product selection tasks, located at @user_8d36cde0/alibaba-1688-trend-product-selection-guoyiyi-ecom. Categorized under business-ops, it requires no API Key. It accepts inputs such as 1688 product details, pricing, sales trends, and supplier information, producing three results: a list of trending product candidates, sourcing risk assessments, and preliminary profit screening. It is suitable for 1688 purchasers, cross-border sellers, and product selection teams to conduct preliminary screening and horizontal comparison. A key limitation is that when no platform connection is configured, it only processes user-provided or exported data; it lacks real-time query capabilities and does not generate market data out of thin air, so data sources must be prepared by the user. As of this writing, official documentation pages and API addresses were inaccessible, and the materials provided did not include installation commands, directory structures, or configuration fields. Therefore, this article only describes general loading methods and does not itemize the differences between Cursor, Codex CLI, and Claude Code, nor does it fabricate input formats or output samples. It is recommended to verify the quality of the output independently before use: as the source is community-based, ratings and download counts displayed on the platform do not imply official certification.
Read MoreEncapsulating Repetitive Team Workflows into Installable, Cross-Tool Agent Skills
This skill targets repeated routines within a team, such as pre-review material preparation and pre-release checklists. It addresses the difficulty of reusing ad-hoc prompts and keeping multiple tools in sync. Inputs may include documents, links, code, or rough process notes; the agent converts them into a reusable workflow based on SKILL.md, completes validation and installation, and can also be shared, optionally, across supported agent tools. It is published by SkillHub community user_3c6cb52e, version 1.0.0, category ai-agent, tag requires_api_key=false. Installation commands and tool-specific differences are not provided in the scraped materials, and the article does not fabricate them; before adoption, verify them against the official page and the generated SKILL.md.
Read MoreBencao Q&A - Millennium Old TCM Doctor: Agent Skill for TCM Consultation with Sources
"Ben Cao Wen Wen - Qian Nian Lao Zhong Yi" is an AI skill for medical consultation based on the original texts of 16 classical TCM classics. Published by user_516d7c35, it is categorized as "professional" on SkillHub, sourced from the "community", version 0.4.0, and requires no API Key. It standardizes the consultation process: following the "Ten Questions Song" for step-by-step inquiries, asking only 2-3 key differential questions per round, and converging to a single syndrome diagnosis. It supports multimodal tongue inspection (tongue color, coating color, coating texture, tongue shape, and moisture). All prescriptions and medical advice must specify the dynasty, author, and source; no prescription is given without a source, and ancient text citations are followed by vernacular explanations. For acute and severe cases, it only provides medical reminders and safety triage. It is to be used with caution by pregnant women, children, and chronic disease patients. It is not applicable for modern medical diagnosis or medication substitution, animal diseases, or non-TCM topics.
Read Moredsh-session-health: DSH Session Status Dashboard
dsh-session-health is a DSH client plugin maintained by kingrenqi-wq. It displays the context usage of the most recent request in a floating widget at the bottom right corner. The plugin provides visual alerts—a yellow border and text—when consecutive similar tools fail or when a task exceeds 300 seconds without reaching a terminal state. It only reads local v4 session logs from ~/.dsh/sessions, with no network access or file writing. Clicking the card reveals evidence such as sampling time and failed sequence numbers. Installation is performed by adding an npm package name, Git URL, or local path via the sidebar, and it takes effect immediately upon activation. This tool is ideal for monitoring context usage and toolchain anomalies during long sessions. Please review the source code and MIT license before installation.
Read MoreConverting Scanned PDFs to Text: Transforming Image-based PDFs into Editable Markdown
This article introduces the community Skill "Scanned PDF to Text" (@user_937442fe/scanned-pdf-extractor, version 3.0.2, category office-efficiency). It targets scanned documents for which pdfplumber, markitdown, or PyMuPDF get_text() all return empty output. By rendering high-resolution PNGs page by page and performing recognition with a configured OCR engine, it outputs structured Markdown. Core capabilities include batch processing; two-stage checkpoint-resumable processing for rendering and OCR; automatic deskew of tilted pages (with three safety gates); correct reading order for two-column documents; and structured table recovery based on PaddleOCR PP-StructureV3. The default vision engine works out of the box with zero installation, with optional local engines paddle and tesseract. The materials do not provide installation commands or configuration fields; the article therefore handles these within the bounds of available facts and provides official links for verification.
Read MoreCompliance Audit Skill: Decomposing Obscure Legal Provisions into Actionable Compliance Points
This article introduces the SkillHub community Skill “Compliance Audit” (@user_3c6cb52e/self-ent-law-compliance-audit, version 2.0.1, category professional). The Skill provides in-depth exploration centered on compliance auditing, translating complex legal provisions into plain language from the perspective of a legal-science writer. It emphasizes practicality and risk prevention, uses the trigger word compliance-audit, and requires no API Key. The article describes its core capabilities and typical usage: submit audit targets such as product features, policy documents, or business processes, and it outputs statute/clause breakdowns along with risk-prevention recommendations. It also outlines applicable scenarios and limitations: it is community-sourced and cannot substitute for advice from a practicing attorney; it has 0 installs and limited real-world usage samples; and provisions must be verified for validity at the time of use. An official link is provided.
Read Moredsh-think-chinese: Preserving Simplified Chinese Reasoning with System Segments and Per-Turn Anchors
dsh-think-chinese is a DSH model reasoning plugin maintained by kikiler114514 (v0.1.0, MIT). It addresses the issue where reasoning tends to slip back into English during long conversations by employing a dual safeguard: an order 10500 system prompt segment and a short user anchor at step 1 of each turn, both locking in Simplified Chinese for thinking and responses. It relies solely on systemPrompt and does not register any tools. This article explains the installation process (adding a local file, two CLI approaches: original/packaged and unpacked), cordis.patch.yml configuration, tracking log verification, and considerations for coexisting with other language plugins, based on the official README.
Read MoreReal-time Market Tracking: Integrating Fragmented A-Share Data into a Single Skill
This article introduces a community skill on SkillHub titled "Real-time Quote Tracking" (@u_8ee34d77/realtime-quote-free, version 1.0.0, category: data-analysis). Targeted at independent developers, enterprise teams, and automated workflows, it allows users to query real-time A-share market data through Chinese interactions, outputting structured results to reduce secondary processing costs. The page indicates requires_api_key=true, meaning an API key must be prepared before invocation. The article explains the core features, enabling methods, typical usage, applicable scenarios, and limitations, and points out that claims in the clues regarding multi-provider fallback and a 3x efficiency boost have not been verified on the official page.
Read MoreWeight Control Assistant: A 5-Step Weight Management Skill Based on Evidence-Based Medicine, Behavioral Science, and Cognitive Science
Weight Control Assistant is a universal Skill based on the SKILL.md format, identified as @user_00c9b356/weightcontrol. Maintained by community publisher user_00c9b356, it is categorized under life-service with version 1.2.0. It breaks weight management down into a five-step closed loop: Diagnosis → Analysis → Suggestion → Tracking → Feedback. The process begins with a 3-minute diagnosis covering BMI, waist circumference, and Emotional Eating Scale (EES) screening. This is followed by a four-layer obstacle analysis examining physiological, environmental, psychological, and habitual factors. Users then select actions from a menu of 12 low-barrier strategies derived from over 10,000 successful cases from the National Weight Control Registry (NWCR). Subsequently, the system facilitates daily weigh-ins, weekly data reviews, and dynamic strategy adjustments. The tool incorporates two-level safety boundaries: a "Red" level mandating immediate medical attention and a "Yellow" level suggesting professional consultation. Its theoretical foundations rely on the 2024 National Health Commission guidelines, the Fogg Behavior Model, and First Principles. The design principles focus on a low entry barrier and changing only one habit at a time. It is explicitly not intended for medical consultation, prescription writing, or extreme diet recommendations. Trigger words include "control weight," "lose weight," "weight management," "slim down," and "weight reduction." `requires_api_key=false`; no API Key is needed. As no unified installation command is provided on the official page, it is recommended to integrate it according to the general SKILL.md format.
Read Moredcc-mcp-skills-creator: A Scaffold for Writing Skills in the dcc-mcp-core Ecosystem
dcc-mcp-skills-creator is an Infrastructure skill maintained by the community publisher user_15292d5a. Currently at version 1.0.0, it is categorized under dev-programming with the tag requires_api_key=false. Designed for the dcc-mcp-core ecosystem, it offers four capabilities—create, validate, scaffold, and review—covering artifacts such as SKILL.md, tools.yaml, scripts, groups, prompts, and skill taxonomy. This assists developers in batch-producing consistent and reviewable Skills, reducing rework caused by missing handwritten fields and inconsistent naming. It is not intended for creating full DCC-MCP adapter repositories; for such requirements, dcc-mcp-creator should be used instead. The boundaries between the two are distinct, and they should not be mixed. As of writing, no direct installation commands are provided on the official page, nor are there differentiated enabling methods for Cursor, Codex CLI, or Claude Code specified; please refer to the official page instructions for details. The official address is https://api.skillhub.cn/user_15292d5a/yjkj-dcc-mcp-skills-creator, and the SkillHub page is https://www.skillhub.cn/skills/user_15292d5a/yjkj-dcc-mcp-skills-creator.
Read Moredsh-tool-exposure: Narrowing Global Plugin Tool Exposure by Agent Presets
**dsh-tool-exposure** is a DSH admin-security plugin maintained by keyblues. It allows configuration of the visibility and callability scope of third-party global tools based on agent presets via the "Tool Exposure" page in Settings. The mechanism works by applying a mask via `ctx.tools.restrict({ deny })` during agent creation or the initial pre-step; hidden tools return **UNKNOWN_TOOL**. Configuration is written to `cordis.patch.yml`, and saving affects only new sessions. It supports wildcard grouping for sources, open/allow role baselines, and group or individual tool overrides. Installation requires entering a GitHub repository address on the plugin page or using `plugin_manager install_bundle`. Requires Node ≥20 and DSH 0.2.1-rc.5+.
Read MoreDashboard Rapid Setup Visual Configuration: Transforming Data Files or URLs into Dashboard Configurations
Addressing the pain points of time-consuming manual dashboard assembly and inconsistent layouts, this article introduces the SkillHub community Skill “数据看板 快速搭建 可视化配置”. Published by user_a10ab3ad, version 1.0.2, and classified as data-analysis, it requires no external API key configuration. It can quickly convert data files or URLs into visual dashboard configurations, supporting batch processing and confidence annotation. The article outlines its three core capabilities and the enablement method based on the universal SKILL.md format, while also candidly noting that the official installation command and usage examples are absent from existing materials and that the official page should prevail. It further cautions that the Skill has a community origin, has an installation count of 0, and that confidence annotation does not replace manual verification of data definitions and metrics.
Read MoreAmap Skill: Delegating Location Search, Weather, and Route Planning to Agents
This article introduces the Amap (Gaode Maps) Skill, an agent skill built on Amap’s Web Service APIs that enables location search, weather queries, and route planning. It states that the skill is published by an enterprise source, maintained by RealQ, released as version 1.0.0, and categorized under life-service, with tags indicating that an API key is required. The article outlines three core capabilities and notes that the official documentation does not provide specific installation commands, so it lists only two prerequisites to confirm before enabling the skill—the API key and the execution environment—while providing the official SkillHub page and the API endpoint. It offers three representative request examples, describes suitable use scenarios, and highlights four cautions: an API key is mandatory, capability boundaries are constrained by the APIs, parameters depend on interface support, and current public data does not yet reflect adoption.
Read MoreFree Video to Text for Any Link: Local, No-Upload Transcription Skill
A community-sourced skill (@user_afd54145/any-video-to-text, version 1.0.0, categorized under data-analysis, tagged requires_api_key=false) that converts video links or local recordings into text transcripts entirely on-device, at no cost and without any uploads. It supports 1,700+ video platforms including Bilibili, YouTube, Douyin, Weibo, Zhihu, Xiaohongshu (RED), and Xigua, as well as local meeting recordings, podcasts, and screen-recorded courses. The backend automatically adapts to the host machine: mlx-whisper runs on Apple silicon, while faster-whisper handles Windows/Linux/Intel Mac, automatically utilizing the NVIDIA GPU when available. It also supports attaching a domain glossary to correct homophone errors in specialized terminology. The article faithfully documents that two official pages failed to be scraped during testing, and that the installation commands and reproducible examples were not verified.
Read MoreFengniao Enterprise Query DSH Plugin: Querying Business and Risk Dimensions Using Agent Tools in Harness
The dsh-fengniao-company-search plugin, maintained by keyblues, integrates with the Riskbird API and registers fengniao_search_company and fengniao_company_dimension in DeepSeek Harness. It first resolves the entid by Chinese company name, then queries 11 dimensions including business registration, shareholders, judicial records, and business abnormalities. Installation is available via npm or GitHub, and the API Key is written into the credential store on the plugin page. Requires DSH >=0.2.1-rc.5 and a Riskbird API Key; each dimension call consumes quota.
Read MoreQuickClass Assignment Converter: Converting Exam Papers and Question Banks into Classroom Assignment JSON
The QuickClass Quiz Converter (MCP server `@user_d101a129/quickclass-quiz-converter`, v1.0.1, community-sourced, categorized under education) converts exam papers, questions, and study materials (Word/PDF/images) into QuickClass-compatible quiz JSON. It limits question types to three categories: SINGLE_CHOICE, MULTIPLE_CHOICE, and TRUE_FALSE, and the page indicates `requires_api_key=false`. The article outlines its trigger boundaries strictly for generating QuickClass quiz JSON (it should not be triggered when merely viewing JSON structures, reverse-exporting, or purely discussing educational content), its input and output types, as well as limitations such as the need to verify each question stem and options before use, and its lack of support for fill-in-the-blank or short-answer questions.
Read MoreReal Estate Viewing Script Coach: Turning a Single Viewing into Trusted Home-Buying Guidance
This skill is designed for real estate agents. It breaks a property showing into four steps: pre-view preparation, route design, structured presentation of key selling points, and objection handling (SPIN). It guides the Agent to complete the showing along a fixed structure, rather than resorting to hard selling. Released by Kunlun Growth (enterprise source), version 1.0.0, categorized under business-ops, with no API key required. The official description only provides the names of the four steps, without scripts templates, route checklists, or SPIN sentence patterns; neither the official site nor the SkillHub page gives a dedicated installation command. It is in the generic SKILL.md format and can be enabled in any Agent tool that supports the protocol. Before implementation, open the official page to verify the details of each step, and use it in accordance with the licensing requirements of your company's internal channels.
Read Moredsh-asset-expert-preset: DSH "Asset Expert" Preset and Enterprise Internet Asset Inventory
The DSH Agent preset bundle maintained by keyblues is built on PTC mode and provides an "Asset Expert" role through cordis.patch.yml. It targets enterprise Internet asset discovery and association graphs, covering ICP filing, FOFA mapping, Fengniao enterprise correlation, Everything searches, and xlsx ledger delivery. Installation is purely declarative and relies on profile-level data sources plus the external asset-collection skill package. It introduces the installation path, prerequisites, typical usage, and acceptance checkpoints.
Read Moredsh-harmony: Managing HarmonyOS Devices and Agent Control in DSH NEXT Desktop
The client-side plugin dsh-harmony (v0.4.1, MIT), maintained by kexijiang, is designed for DSH NEXT 2.0.21-next on Windows x64. Depending on Better Sidebar to register a "Harmony Device" tab, it supports switching between sidebar and pinned floating screen casting, theme synchronization, and a file tree with a UI inspector. Through harmony_control, it provides agents with enumeration, observation, touch control, and screenshot attachment capabilities. Installation requires downloading dsh-plugin-harmony.tgz from Releases and enabling it within NEXT; configuration parameters such as hdcPath and cwd are detailed in cordis.patch.yml. The plugin operates with DSH process privileges; please review the source code and license before use.
Read MoreRude Quant Investment Research: Strategies and Backtesting Skill for Stocks, Convertible Bonds, and Funds
Lude Quant Research is an AI-powered investment research product developed by Lude Network. Identified as @lude/lude-quant (version 0.0.1, category: Professional) and published by Lude Network, its official URL is https://api.skillhub.cn/lude/lude-quant. It is designed for quantitative strategy design, custom factor expressions, target pool management, historical target selection, and backtesting validation across stocks, convertible bonds, and funds. It is ideal for activation when users explicitly request writing or modifying Lude strategies, building factors or target pools, or saving strategies and analyzing backtests. This article outlines its core capabilities, the scope of available resources, and typical request methods, while noting that installation commands, configuration fields, ratings, download counts, and runtime permissions should be based on real-time information from the official page.
Read MoreFineReport 11.0 - FVS Generation Assistant: The Cockpit Development Pipeline from Requirements to .fvs
FineReport 11.0-FVS Generation Assistant is a full-chain skill designed for FineReport 11.0 dashboard and data screen development. Maintained by the SkillHub community publisher user_b6ce042c, it is categorized under data-analysis (Version 1.5.0). It bridges three stages: organizing customer requirements, generating high-fidelity prototypes, and converting design drafts into .fvs files. Upon receiving raw requirements, it outputs a structured requirement outline, a list of key metrics, and calculation formulas along with data definitions. Based on the outline, it intelligently matches chart types, plans layouts, and generates multiple high-fidelity PNG dashboard prototypes, supporting multi-page navigation and style customization. By connecting to a local FineReport 11.0 project, it converts design images into editable .fvs files accessible in the designer; officially, it claims a visual restoration accuracy of over 95% with fully editable components. These three functions can be executed as a continuous workflow or invoked independently, making it suitable for rapid prototyping and delivery of data screens, management dashboards, and business kanbans. To use it, mount the skill within a host that supports SKILL.md; no unified installation command is provided on the page. It is triggered by keywords in the conversation such as "dashboard," "big screen," or "kanban," or when the user needs to organize metric systems, generate prototypes, or convert design drafts (or Lanhu cutouts) into .fvs files. The article also notes that this page only verifies the three core capabilities, and data implementation and project delivery details need to be confirmed with the publisher.
Read Moredsh-usage-hub: Provider Usage Dashboard Adaptive to Current Model Switching
dsh-usage-hub is a DeepSeek Harness plugin maintained by KevenUnderwood. It binds to session modelSelection, automatically switching query targets when models are changed, and displays quota windows, account balances, and session as well as cross-session token summaries. It includes built-in adapters for multiple providers and independent provider catalog pages, with credentials showing only reference names. Installation requires Node.js ≥ 20; the recommended command is `dsh plugin --profile web add dsh-usage-hub`, followed by restarting dsh web. Licensed under MIT, categorized under admin-security.
Read MoreTop 5 Wall Street Daily: US Stock Market Sector Board Generator
Wall Street Daily Top 5 is a community-published "Professional" category Skill located at @user_c8b52036/huaerjietop5, tagged with `requires_api_key=false`. It generates a dark, professional-style HTML dashboard—ready to open directly in a browser—by synthesizing daily popular US stock sectors, industry/theme rankings, capital flows, and in-depth analysis. The dashboard includes a market overview bar, Top 5 sector cards, and collapsible five-dimension deep analysis cards, along with a sector rotation heatmap, capital flow Sankey diagram, and an RSI thermometer. Data is based on public market information and can be entered manually or scraped online; popularity scores are calculated by the generator using a specific formula rather than relying on real-time APIs. The output includes field-level confidence labels and a disclaimer. It is suitable for individuals needing daily sector reviews, industry/theme rankings, and capital flow analysis. The official page was inaccessible, and ratings and download counts remain unverified; hence, they are not cited in this article.
Read MoreOPC Legal Counsel: Triage Skills for Legal Services for One Person Companies
OPC Legal Counsel (@user_55833ee9/opc-legal-counsel, v1.0.2) is a community-sourced professional skill on SkillHub designed for legal business triage tailored to one-person companies and small-to-micro enterprises. It identifies primary conflicts and associated risks, and provides action priorities and escalation boundaries. For any inquiries regarding current statutes, validity status, deadlines, tax rates, filing registries, local policies, administrative procedures, or platform rules, it generates structured retrieval tasks for verification by legal databases, MCPs, or official sources, ensuring it does not substitute built-in knowledge for the latest regulations. The official description explicitly states that it is not intended to replace formal legal advice, in-depth contract annotation, litigation documents, trademark applications, patent analysis, or pure legal research. It is tagged as requires_api_key=false, with a recorded rating of 100,000 and 949 downloads.
Read More