dsh-speech-to-text-gigaam: DSH Local Russian GigaAM Speech Recognition
The DSH plugin `dsh-speech-to-text-gigaam` (v0.4.0), maintained by kelvmg, is based on Sber GigaAM v3 E2E RNNT. It performs Russian dictation locally and outputs text with punctuation and capitalization, ensuring that audio and transcripts do not leave the host machine. To use it, install the experimental voice input bundle in DSH, add the GitHub repository via the Plugins section, and enable it. Then, select "GigaAM v3 · Russian" in the voice settings and click "Download and prepare." Version 0.4.0 supports one-click isolation for Python, PyTorch, and model preparation on Apple Silicon macOS 13 and later. It requires approximately 4GB of disk space and 2GB of RAM.
Read MoreSector Research Report Interpretation: Structured Report Analysis Skill for Industry and Thematic Directions
This article introduces the community Skill "Sector Research Report Interpretation", located at @user_d196d58d/investoday-sector-research-interpretation-v2, version 1.3.0, published by user_d196d58d, categorized under "professional" in SkillHub, with the tag requires_api_key=false. This Skill is designed to interpret research reports across industry, sector, concept, and thematic directions, focusing on sell-side consensus, core logic, opportunities and risks, and representative directions. Based on the Jinri Touzi financial data API, it automatically identifies industry or concept entities and outputs structured sector research report interpretations. Trigger words include sector research reports, industry views, track consensus, theme divergence, and key directions to watch. The article explains its positioning, verified capabilities, official address, and trigger-based usage, and notes that installation commands and configuration items should refer to the official page, that the data scope is subject to the Jinri Touzi API constraints, and that it does not constitute a return promise.
Read MoreKunlun Growth · C# Development Expert: Skill for .NET 8 Application Development
Kunlun Growth · C# Development Expert is an agent skill compatible with the general SKILL.md format, released by Kunlun Growth. Identified as an enterprise source with version 1.1.0, it locks .NET 8, ASP.NET Core, and Blazor as the technical baseline. This enables development agents to generate and modify applications according to a set of conventions when handling C# requirements, without the need to repeatedly select or justify the tech stack. The metadata tag `requires_api_key=false` indicates that no additional API Key configuration is required for invocation. This article introduces what it is, its verified core features, installation and enablement methods, general usage workflows, as well as applicable scenarios and limitations. It should be noted that there is a discrepancy between the clues (rating 99,545, downloads 875) and the retrieved data (rating 100,000, downloads 897); the retrieved data is considered authoritative for this analysis.
Read MoreVoice Workshop (dsh-plugin-voice-studio): Integrating TTS, Covers, and QQ Voice into the DSH Plugin
Voice Studio is a DeepSeek Harness plugin maintained by keduoli-lovely, categorized under model inference. It integrates six TTS pipelines including IndexTTS, GPT-SoVITS, and Sour AI, MiniMax, MiMo, OpenAI-compatible APIs, a reference voice library, an emotion chain (capable of connecting to the DSH emotion plugin or using the built-in dictionary), RVC song covers, as well as integration with the qq_send_voice function via qq-bridge and random voice replies, all consolidated within Settings → Voice Studio. The Agent utilizes tools such as voice_generate and song_cover. The installation command is dsh plugin --profile web add github:keduoli-lovely/dsh-plugin-voice-studio. Requirements include DSH ≥0.1.7 and Node ≥18; if optional services are missing, only the corresponding functionalities will be unavailable.
Read MoreInitiating a Proposal Critique: Conducting a Branch-by-Branch Stress Test on the Proposal Prototype
"Initiate Plan Interrogation" is a review-oriented Skill published by Kunlun Growth on SkillHub. Version 1.0.0, it is categorized under business-ops, originates from an enterprise source, and is tagged with requires_api_key=false. The page boasts a rating of 100,000 and 951 downloads, with 0 installations. Targeting users with existing plans or preliminary decision concepts, it launches a round of pressure-testing during the formulation phase. This process examines the plan branch by branch, provides recommended answers, and shifts the review process from holistic evaluation to item-by-item judgment. This article outlines the skill's core capabilities and boundaries, as well as the configuration details regarding the lack of an API Key requirement. It notes that official documentation does not provide installation commands, illustrates typical usage through an input example constructed based on the description, and concludes with information on the target audience, input completeness requirements, and the official URL.
Read MoreDragon and Tiger List - CICC Wealth: Query for Stocks with Unusual Trading on Shanghai and Shenzhen Dragon and Tiger Lists
Introducing the “Dragon-Tiger List - CICC Wealth” Skill published on SkillHub: produced by CICC Wealth Securities, version 1.0.2, categorized under professional, providing query of stocks listed on the Dragon-Tiger List with unusual movements on the Shanghai and Shenzhen stock exchanges, with search by seat and by date. The article outlines the core features, the prerequisite for API Key configuration, typical use cases (by date, by seat), and data scope limitations, and provides the official address.
Read MoreSocialDataX Trend Analysis: Access to the Skill for Xiaohongshu Trend Observation
SocialDataX Trend Analysis (@user_825d9e7e/socialdatax-trend-insights-v3, v1.0.5, community-sourced, categorized as data-analysis) targets RedNote / XHS content scenarios, covering trend radar, trend insights, trend analysis, hotspot observation, content direction assessment, trend clue induction, and marketing inspiration curation. The tag `requires_api_key=true` indicates that an API Key must be prepared before invocation. The official page retrieval failed this time; consequently, the installation command, directory, and configuration fields were not verified. This article does not fabricate information and provides only verified data (Rating: 100000, Downloads: 907, Installs: 0), while suggesting that community-sourced content requires small-scale validation. Information is current as of 2026-11-04.
Read MoreThe Mind Mapping Study Method: Visualizing Textbook Knowledge
# Introduction The most common problem encountered by primary and secondary school students during revision is that "the book becomes thicker the more they read." Knowledge points are scattered across textbooks, examples, and class notes, leaving students reliant on flipping through pages to re-read material before exams. The approach of this Skill is to consolidate scattered knowledge into a single map—first defining a central theme, then expanding branches hierarchically, and mapping out the relationships between concepts, rather than simply transcribing original sentences. The following is a general-purpose SKILL.md Skill designed for this scenario. ## What is this? The "Mind Map Learning Method" is an enterprise-sourced Skill on SkillHub, located at `@kunlungrowth/edu-mindmap`. It is published by Kunlun Growth, with version 1.0.0, under the education category. In one sentence: it uses mind maps to organize knowledge and...
Read Moredsh-phoenix (Phoenix): Preserving DSH and Configuration during Plugin Failure
An introduction to the DSH plugin dsh-phoenix maintained by KDreamelf: it intercepts crashes when a plugin causes the process to exit or the UI to freeze, attributes the offending plugin, and isolates only that line. It provides suspend / give a second chance / dismiss and one-time revive modes, without touching the in-settings providers or other plugins. Includes the official installation command, restore usage, desktop-side limitations, and applicable scenarios.
Read More"Happy Birthday" Skill: Using macOS LaunchAgent to Automate WeChat Birthday Blessings on a Schedule
The "Happy Birthday" skill is a community-developed macOS automation Skill (located at @user_9406e227/send-wechat-birthday, version 1.0.0, categorized as dev-programming, and requiring no API key). After inputting a WeChat ID and a sending time, a foreground GUI automation script is triggered on schedule via macOS LaunchAgent on that day. This script automatically launches the PC version of WeChat and sends the message "Happy Birthday." The article covers core features, coordinate-based installation, typical workflows, and usage limitations, including that it is exclusive to macOS, relies on desktop WeChat, requires an active desktop session for GUI automation, may truncate official notes, and currently has zero installs. It is recommended to review the complete instructions on the official page before use.
Read MoreData Curator: Data Cleaning Assistant
Introducing the Data Curator Skill on SkillHub. Published by Zhiyu Technology, this skill is listed under the source category "enterprise", falls within the "professional" classification, and is at version 1.0.0. The page shows a rating of 100000, 54 downloads, and 0 installations, with requires_api_key=false. According to the official description, it is designed to help white-collar workers handle repetitive tasks related to scientific research, enabling them to quickly obtain structured and reusable work deliverables. The page does not provide installation commands, directory paths, or configuration fields, nor does it list a detailed set of sub-features or reproducible examples. Its actual capabilities must be verified through real-world usage. The article presents verified metadata, a general scientific research data processing workflow based on the introduction, as well as applicable scenarios and important considerations.
Read Moredsh-ponytail: Minimal Code Rules and Desktop Capabilities for Integrating Ponytail 5 into DSH
The dsh-ponytail client plugin from KB1SLN-Labs is an unofficial port of Ponytail 5.3.3's minimalist code rules to DeepSeek Harness, retaining upstream rules and skills such as review/audit, while providing a settings page, mode selector, Composer status bar, and session usage dashboard. Compatible with DSH >=0.2.1-rc.3 and <0.3.0, currently in Beta. The article covers Desktop and CLI installation commands, the /ponytail series usage, mode and session override precedence, as well as notes on migrating from other adapters and reviewing the source code before installation.
Read MoreForward Writing Engine: Stylized Skill for Chinese Writing Tasks
The Qianjin Writer Engine (ID: @user_d3228c23/qianjin-writer, version 2.10.0, category: content-creation, community-sourced, no API key required) consolidates multiple Chinese writing styles into a single Agent Skill — including People's Daily / Xinhua News Agency prose, viral social media structures, international wire service and domestic press release paradigms, non-agency special press releases, personal voice, marketing copy, and "de-AI-ifying" rewriting — positioning itself with core editorial stances aligned to Xinhua News Agency and People's Daily as domestic media authorities. The article examines developers' pain points regarding style switching and stance consistency, outlines its positioning and verified capabilities, provides official addresses and coordinates, and identifies gaps in publicly available documentation — specifically the absence of installation commands, usage examples, and implementation details for its "self-evolving" mechanism. With zero installations recorded, its actual usage scale remains unverifiable. The author recommends checking the official page for distribution details and version information before integrating, while noting that community-sourced content requires manual review and that preset editorial stances should be evaluated against your business's editorial stance.
Read Moredsh-linear: Directly Connect to Linear GraphQL using 26 Tools in DSH
The `@kb1slnlabs/dsh-linear` package maintained by KB1SLN-Labs registers 26 `linear_*` tools for DeepSeek Harness, connecting directly to `https://api.linear.app/graphql` and using a personal API key to manage Issues, comments, relations, teams, labels, projects, cycles, and users. Version 0.2.x removed the MCP bridge that existed in 0.1.0. The article covers the scope of the tools, cursor-based pagination, the configuration priority among the `apiKey` parameter, the `LINEAR_API_KEY` environment variable, and the `key` file, as well as the official installation commands and self-check usage for Desktop and CLI (web/desktop profile).
Read MoreStrategic Standards Inspector for Vocational Education: A Three-Dimensional Assessment of Coverage, Policy Alignment, and Strategic Value Based on Document *Jiao Zhi Cheng* [2026] No. 1
A strategic, standards-based health check skill for vocational education teaching products at the secondary, higher, and vocational bachelor’s levels. Anchored by the policy framework in Opinions on Deepening the Reform of Key Elements in Vocational Education Teaching (Jiao Zhi Cheng [2026] No. 1), it performs a three-dimensional assessment—coverage, policy alignment, and strategic value—across teacher workbenches, courses, practical training platforms, question banks, and textbook/ teaching resources, and outputs a prioritized roadmap. Trigger terms for registration include: vocational education standards health check; five-element alignment; program construction diagnosis; course standards benchmarking; practical training platform assessment; industry–education integration verification; 1+X alignment; alignment with the vocational education college entrance exam; and school-running quality health check. Published by user_0898a299, version 1.1.19, source: community, category: professional, requires_api_key=false (API key not required). No installation commands or configuration fields are provided in the official materials; this article explains how to enable the skill in accordance with general SKILL.md conventions, and reminds users to dynamically verify community-sourced data. The conclusions do not replace close reading of the original policy documents or expert review.
Read MoreRuijin Gastroenterology Case Analysis Review: Evidence-based Clinical Teaching in DOCX
**Preface** Grading clinical pharmacy case analysis assignments is a laborious task for instructors. They must review each paper individually: first checking for missing basic requirements, then assessing the clarity of the case summary, followed by verifying the identification of pharmaceutical events and the depth of discussion in the analysis and summary, and finally confirming whether the references genuinely cite relevant guidelines. Grading a single paper can take over ten minutes, and with dozens of papers piled up, the greatest challenge is ensuring that every piece of feedback lands at a specific location rather than appearing as a generic comment. The "Ruijin Gastroenterology Case Analysis Grading" Skill introduced below is designed to handle exactly this issue. It codifies the instructor's actual grading experience into a set of scoring rules and inserts feedback directly into Word documents in the form of native comment bubbles. The evaluation content is integrated into the corresponding comments rather than forming a separate section. **What is this?** In one sentence: An AI grading skill designed for clinical pharmacists' gastroenterology case analysis assignments.
Read MoreDX-S2-01 Department Drug Catalog Management: Department Drug Catalog Service for Doctors and Pharmaceutical Companies
DX-S2-01 Department Drug Catalog Management is an enterprise-level skill released by the Baitong AI Service Platform, categorized as "professional," version 2.2.5, with the coordinates @baitongai/dx-s2-01-ke-shi-yao-pin-mu-lu-guan-li. It provides niche services in department drug catalog management for doctors and pharmaceutical companies. It belongs to the Doctor Supply Side Blind Spot (S2) direction of Hospital Pharmacy Administration, within the cross-matrix P4L3, and is officially marked as requiring no API Key. This article introduces its positioning, verifiable core information, installation and activation methods, typical usage, applicable scenarios, and precautions. It should be noted that as neither the official page nor the SkillHub page was accessible at the time of writing, no unverified descriptions have been made regarding installation commands, configuration fields, input/output formats, or specific functional details. regarding public data, the rating is 100,000, with 887 downloads and 0 installations, indicating that there is currently no public installation data, which should be included in your evaluation when selecting. It is recommended to reconfirm installation methods and configuration requirements via the official SkillHub page before use, and to read the accompanying SKILL.md to understand specific invocation details.
Read Moredsh-plugin-ntfyx: Connecting DSH Task Progress and Mobile Confirmations to Ntfyx
dsh-plugin-ntfyx 0.1.0 is a DSH community client plugin maintained by Ntfyx, which launches the local ntfyx mcp via cordis.patch.yml and registers the ntfyx skill, providing MCP tools such as protected notifications, mobile inquiries, and claim-based responses. It is compatible with DSH 0.2.2-alpha.2 and Node 24.9+. Installation requires pairing the Ntfyx CLI first, then using dsh plugin add dsh-plugin-ntfyx@0.1.0. Completion notifications and confirmations rely on agent tool invocation and do not replace DSH permission policies.
Read MoreViral Short Video Production Line: Real-Person Talking Head Edition – Converting Talking Head Footage into Vertical Final Videos
This is a paid automatic voiceover video editing production line released by Beijing Ruixin Technology (version 1.1.4, categorized as design-media). It processes on-camera or voiceover videos locally into unified 9:16 vertical (1080×1920) finished outputs. Modules include filler-word removal, dead-air/silence removal, audio enhancement (noise reduction + loudness normalization), burned-in subtitles (96 px Source Han Sans, single-line greedy fill, no punctuation + pause-aware line breaks), picture-in-picture product demonstrations, and cover images (white background with gold-text titles). It is suitable for voiceover editing, short-video production, e-commerce videos, knowledge-based talk videos, batch video processing, disfluency removal, automatic subtitles, picture-in-picture demos, and automatic finished-video generation. The official documentation does not provide directly copyable installation commands, directory conventions, or configuration-file fields, and does not describe differences in enabling it across Cursor, Codex CLI, or Claude Code; this article therefore makes no assumptions. Key notes include: outputs are 9:16 vertical only; subtitles are burned in rather than on a separate subtitle track; covers use white background with gold text; and the paid version requires confirming licensing/authorization.
Read MoreCover Image Generator (Baoyu Cover Image): Delegating Article Cover Parameters to Skill
Baoyu Cover Image is a cover generation Skill utilizing the universal SKILL.md format, identified as `@jimliu/baoyu-cover-image`. Published by jimliu (Jim Liu 宝玉) on ClawHub, it is categorized under design-media (Version 1.117.5) and declares `requires_api_key=false`. The skill breaks down cover creation into five dimensions: Type, Color Palette, Rendering, Text, and Mood, with an additional Font dimension. It offers 11 color palettes and 7 rendering styles, with specific values and default settings defined for each dimension. Supported aspect ratios include 16:9 (default), 2.35:1, 4:3, 3:2, 1:1, and 3:4. Installation can be performed via `openclaw skills install @jimliu/baoyu-cover-image` or `npx skills`. The image generation backend is resolved by priority: the backend explicitly named in the current request, `preferred_image_backend` in `EXTEND.md`, runtime native `imagegen` or `GenerateImage`, Codex CLI, other native tools, or non-native `baoyu-image-gen`. The SKILL.md enforces two strict constraints: SVG, HTML, or canvas must not be used to substitute bitmap images, and text overlay patching on already generated bitmaps is prohibited. Before generating an image, the complete prompt must be written to a separate file under `prompts/` to serve as a reproducible record. Confirmation of dimensions, language, and backend is required before generation by default, unless explicitly waived or set to `quick_mode`.
Read MoreAI Official Document Writing: Embedding the Structure of Public Sector Documents into Prompts
A Chinese Agent Skill designed for official document writing for government agencies, enterprises, and public institutions. Located at `@user_ef9e183a/ai-writing-gongwen-xiezuo` (Version 1.0.0), it is released by community developer user_ef9e183a and categorized under content-creation; no API Key is required. The skill incorporates built-in formatting constraints and libraries of common terms and phrases, covering the drafting of initial versions for document types such as notices, reports, requests, and summaries. The article outlines the pain point it addresses (the need to repeatedly specify formatting in every conversation), its two core capabilities, and installation instructions: as no official installation command is provided, it must be placed into the client's skills directory following general SKILL.md conventions; loading mechanisms for Cursor, Codex CLI, and Claude Code are not specified in official documentation. It is also noted that the SkillHub page displays a rating of 100,000, 1,028 downloads, and 0 installations, indicating a lack of community validation; therefore, rules for document applicability, policy data, and proper nouns still require manual verification.
Read MoreEquipment Inquiry Expert: Entrusting the Inquiry, Price Comparison, and Compliance Review of Educational Informatization Equipment to a Skill
# **Foreword** Professionals involved in education informatization procurement frequently face a type of repetitive labor: upon receiving a list of equipment names or specifications, they must search various public channels—such as government procurement networks, JD.com, and electronic marketplaces—to find equivalent models. This process requires manually compiling comparison tables and ultimately assessing whether specific parameters pose risks of exclusivity. Given that these workflows involve fixed steps and open sources, they are well-suited for Agent Skills. The "Equipment Inquiry Specialist" (located at `@user_924b0a5e/yrns-equipment-inquiry`) introduced below is designed for exactly this scenario. It does not teach operating systems or maintenance; it does one thing only: based on the equipment names, specifications, or descriptions provided by the user, it searches public information to find matching or equivalent devices, outputting a parameter comparison table along with a recommended ranking. #
Read Moredsh-task-sfx: Three sets of mutually non-interfering task sound effects
kanmanJO's DSH plugin dsh-task-sfx provides three independent sound effect pools for Web sessions: Pool A plays a random sound at the end of each reply round; Pool B can be enabled to play a reminder at a configurable interval when no new activity occurs; Pool C triggers a reminder for approval, ask_user_question, or exit_plan_mode events. Settings, intervals, and audio files are persisted in the task-sfx directory under the DSH home. The plugin supports npm or GitHub installation and has no third-party dependencies. On first use, you need to preview the sounds in the panel to satisfy the browser's autoplay policy.
Read MoreBook Shortage Savior: Turning 'All-Platform Book Search' into a Single Query
*Book Famine Saver* (@user_66f29720/a-savior-of-book-drought, v1.0.0, community, knowledge-management) is a novel recommendation Skill. By taking input of desired genres or preferences, it searches across all platforms to generate recommendation plans, outputting book titles, ratings, synopses, reader reviews, and reading links. It covers two source categories: web novels (Qidian, Tomato, Qimao) and traditional literature (Douban, Zhihu book lists). Trigger words include “推荐小说”, “书荒”, “找小说”, “想看小说”, novel, book, “推书”, “求书单”, and “小说推荐”. It is configured with requires_api_key=false, meaning no additional API Key is needed. The article outlines its positioning, core capabilities, coverage, activation path, and typical use cases, while highlighting usage limitations such as the inability to directly compare cross-platform ratings, the small sample size of Zhihu book lists, and its current installation count of 0.
Read MoreGlobal Rules Panel: Write persistent rules in the DSH sidebar once, and they are automatically included in every request.
Maintained by kanmanJO, `dsh-global-rules-panel` provides a global rules panel within the DeepSeek Harness sidebar. Rules are saved to `<DSH home>/global-rules/rules.json` and injected via the `systemPrompt` section of the root context, covering the main session, sub-agents, and new sessions. It supports a 20,000-character limit, instant saving, a disable toggle, and can coexist with `AGENTS.md`. Recommended installation: `dsh plugin --profile <profile> add dsh-global-rules-panel`. MIT License. Requires `dsh>=0.2.1-rc.3`.
Read More