dsh-entry-organizer: Drag-and-drop sorting and cross-region movement of sidebar and tabs
dsh-entry-organizer is a plugin for the DSH client maintained by xishuai-zero. It allows adjusting the order of entries in the left global panel and tabs at the top of sessions via drag-and-drop or buttons on the settings page, and can move sidebar items to the tab bar to open panels as real Views. The order and move state are stored in localStorage, persist across restarts, and synchronize across multiple windows via storage. It has zero dependencies, is licensed under MIT, and requires host version >=0.2.1-rc.5. Installation is available via npm, GitHub tags, or Release tarballs. Note the documented boundary limitations: tabs in the overflow menu cannot be dragged, state is local-only, and it depends on the host's DOM contract.
Read MoreClawhub: A General Skill Pack for AI Agents
Published on SkillHub at @user_3c6cb52e/clawhub, this skill is sourced from the community, categorized as ai-agent, and is currently on version 1.0.0. Its description reads: "Specialized skills tailored to Clawhub, designed to help AI Agents complete relevant tasks efficiently." The page indicates requires_api_key=false, with a rating of 100000, 833 downloads, and 0 installations. Since the SKILL.md file content was missing from the retrieved data, this article did not fabricate feature lists, installation commands, or examples. Instead, it only outlines the general structure of SKILL.md-type skills and points out issues such as the zero installation count, the vague description, and discrepancies between the rating/download counts in the retrieved data and the official page. Before implementation, users should request the SKILL.md file from the publisher to verify its capabilities and usage.
Read MoreIMA Diary Append: Directly append content to the end of IMA Diary notes
"ima diary append" (@user_5ee82744/ima-diary-append) is a community Skill based on the SKILL.md format, published by user_5ee82744, version 1.0.0, categorized under knowledge-management, and carries the tag requires_api_key=true. It addresses the problem of needing to repeatedly search by title and easily writing content to the wrong location when adding entries to IMA diary notes: on first use, it automatically discovers or creates a note named "日记" (Diary) and caches the note id locally; after that, it directly uses the cached id to append, without further searching. When appending, it automatically writes a timestamp, and content is only added, not modified. The trigger is keyword-driven, activating when phrases such as "日记" (diary) or "记录" (record) are recognized. The official example: when a user says "记录这个skill挺好用的" ("this skill is pretty useful"), it can append the timestamp and this sentence to the diary note. Note that an API Key is required; note positioning depends on the fixed name "日记"; this Skill is a personal customization based on ima-skill, and long-term maintenance depends on the publisher. The installation command and client differences mentioned in the article were not verified because the official page was inaccessible, and no content was fabricated.
Read MoreEnterprise Subsidy Policy Matching Engine: Automatic Matching of Government Subsidies Based on Enterprise Profiles
The Enterprise Subsidy Policy Matching Engine is a "business-ops" Skill published on the SkillHub community (v1.0.3, by user_84cc09f2, no API Key required). By analyzing enterprise profiles, it automatically matches policies across six major categories: technology innovation, talent, industrial support, energy conservation and environmental protection, digital transformation, and international trade, generating personalized application plans. This article details its core features, keyword trigger mechanisms, installation coordinates and activation methods, and typical use cases. It also clarifies that it is not applicable to scenarios such as personal subsidies, housing provident fund withdrawals, or individual social security inquiries.
Read Moredsh-archive-cleaner-panel: Web-side thorough cleanup of archived conversations in the default workspace
dsh-archive-cleaner-panel is a Web plugin for the DSH client maintained by xinzhi-liu. It provides an "Archive Cleaner" panel in the sidebar that lists only archived conversations within the default workspace. The plugin supports permanently deleting session logs, projection caches, workspace archive/pin/ownership records, and derived agent session logs after manual selection and secondary confirmation. The Host side validates the archive status and workspace ownership, and the deletion is irrevocable. It applies to Web profiles and requires DSH >=0.2.1-rc.3 <0.3.0 and Node.js >=22. To install: dsh plugin --profile web add github:xinzhi-liu/dsh-archive-cleaner-panel. MIT License.
Read MoreWeChat Viral Article Writing Assistant: Hot Topic Drafting with Dual AI Fact-Checking
A technical introduction to the SkillHub community skill "WeChat Official Account Viral Article Writing Assistant (Trending Topics & Dual-AI Verification)." Users input a trending topic to generate an in-depth article exceeding 1,000 words. The title includes relevant keywords, and the body naturally integrates the topic. The workflow includes dual-AI fact-checking, automatic image insertion, and full-platform publishing, supporting WeChat Official Accounts, Zhihu, and Toutiao. This article explains the specific problems it solves, four verified core capabilities, and the activation method under the universal SKILL.md format (`requires_api_key=false`). It also outlines the reproducible workflow based on the official description. Installation commands and configuration fields are notably absent from the provided material and are honestly marked as such in the text to avoid misinformation. Additionally, it lists three specific limitations: AI-generated content requires human review, monetization potential does not equate to actual traffic, and publishing API details require real-world testing.
Read Moredsh-any-rollback: Restoring the workspace (including command artifacts) during conversation rollback
dsh-any-rollback is a DSH workflow plugin maintained by xinstart-1 that can roll back to any user message within the current window and optionally restore workspace files, including command artifacts such as ffmpeg and build scripts. The plugin backs up data using turn-boundary manifests and content summaries, and moves files to an isolation zone instead of deleting them during rollback. It covers installation (via Release tarball and dsh.profile.bundles), the /rewind and snapshot cleanup commands, configuration options, and the scope of application.
Read MoreOptimize AI Reply
Introducing the DeepSeek Harness community plugin `dsh-plugin-optimize-ai-reply` (XINGMENG2333, Workflow). Powered by seven tools, the plugin performs local density scanning, task-driven rewriting, five-layer verification (factuality, strength, and safety), and closed-loop rescanning; if it fails acceptance, it falls back to the original text. It includes official installation commands, Cordis configuration, an interface master switch, and typical conversational usage examples. It emphasizes zero network scanning, no reliance on third-party AI detection APIs, MIT licensing, and the important note to review the source code before installation.
Read MoreBusiness Model Expert: A Skill for Designing and Redesigning Business Models for Founders
**Business Model Expert**, developed by Zhihe Island Group, is designed for **founders** and**operators**. It deconstructs "the fundamental logic behind a business's viability" into three**verifiable** frameworks:**three-party intent**,**seven clusters of 108 elements**, and**five-step design SOP**. The process sequentially performs an**intentional diagnosis**, applies elements, assesses combination conflicts, and aligns with the eight mainstream models. This yields a practical action plan featuring**element combination cards**, a**four-step flywheel assessment**, and an**A/B testing KPI dashboard**. Conclusions conclude only in three states—**Satisfied**,**Unsatisfied**, or**To Be Supplemented**. Deliverables are written in accordance with the set methodology, clearly indicating**evidentiary rigor**. The system also issues referral notices for topics regarding equity, financing, and IPO. This article introduces its positioning, core functions, access points, typical usage, and applicable boundaries.
Read MoreKingdee Cloud Galaxy Expense Assistant: Employees Can Submit Reimbursements Without Logging Into Kingdee
The Kingdee Cloud Reimbursement Assistant, developed by the Little Beaver Studio (Version 2.3.5, Category: ai-agent), enables employees to submit invoices via chat without logging into Kingdee. Leveraging pure WebAPIs, the agent automates the entire workflow: generating an "Expense Application Form," pushing it to create a "Reimbursement Form," associating invoice receipt information and uploading attachments, and finally submitting the request for approval. The travel expense workflow follows an isostructural process. Key features include built-in validation for invoice clarity and title, separation of invoice receipt information versus attachments (featuring dual-calculation warnings for itineraries), hard interception of cross-organization invoice association, and visibility isolation for receipt pool accounts. The system enforces four strict submission gates (zero receipt rows, missing invoice cloud serial number, cross-organization issues, or total invoice amount less than the reimbursement amount) with no bypass parameters available. Additionally, it supports adjusting the reimbursement amount downwards to match invoice amounts (fit), server-side invoice uploading with automatic receipt creation, a unified login entry, and pre-submission checks. This tool is suitable for expense reimbursement, travel reimbursement, employee reimbursement, invoice association, and approval submission scenarios within Kingdee Cloud (K3 Cloud). API Key configuration is required.
Read Moredsh-dfy-mirror: Using scrcpy Windows in DSH for Android Screen Mirroring
dsh-dfy-mirror (xiaozqi/dsh-dfy-mirror) is a DSH client plugin on SkillHub that adds Android screen mirroring to Harness. The screen is displayed via a native scrcpy window, and the plugin provides three access points: a floating window, a desktop window, and a settings page. The documentation indicates support for Windows and Android. This summary is based on the directory listing description and installation instructions. As the GitHub README could not be retrieved and the repository page was inaccessible at the time, installation commands and configuration items have not been included to avoid speculation. This plugin is suitable for intelligent agent development scenarios on Windows using DSH where visual verification on a physical device is required. Before installation, users should review the source code and license, and note that the plugin runs with the privileges of the dsh process.
Read MoreFinancial Reconciliation SKILL: Delegating Multi-Source Transaction Reconciliation to Agents
# Introduction Financial reconciliation is a highly repetitive task where errors are intolerable. Bank statements, payment channel records, platform settlement sheets, and order details are scattered across multiple Excel/CSV files with inconsistent formats. Manual verification relies on personnel familiar with the business to match entries one by one, flag discrepancies, and write explanations. Once the data volume increases or ownership is transferred, both costs and error rates escalate significantly. The "Financial Reconciliation SKILL" is an Agent Skill designed to address this problem. Its positioning, capabilities, and usage are introduced below. # What is this The Financial Reconciliation SKILL is a general-purpose financial reconciliation Skill, with coordinates @user_7a244bf0/3b5a7c9d-1e2f-4a8b-9c0d-1e2f3a4b5c6d, version 1.0.0, developed by the community.
Read MoreBlender: The bpy Python Bridge for Command-Line Control
Blender is a dev-programming skill on SkillHub, published by RealQ. It utilizes Flue as the Python bridge layer to connect directly with Blender's bpy, driving Blender from the shell without requiring an MCP server or API key. This article verifies its positioning, core capabilities, and usage prerequisites, and notes that the installation command could not be verified on the official page.
Read MoreKōngwéi Academic Writing Assistant: Integrating Paper Drafting, Polishing, Translation, and Literature Retrieval into a Single Skill
Kangwei - Academic Writing Assistant is an Agent Skill designed for scientific research writing, located at @user_aa2ab091/academic-writingassistant. It is published by user_aa2ab091, sourced from community, versioned at 1.0.2, categorized under professional, and tagged with requires_api_key=false, meaning no API key is required for activation. It integrates three workflows—drafting academic manuscripts, polishing expression and translating, and finding supporting literature—into a single entry point. The Skill automatically detects user intent and routes to corresponding sub-workflows, eliminating the need for users to manually specify which path to follow. Its writing principles are based on Nature style, with literature search scope limited to CNS. Results can be exported in citation format, and its profile indicates it primarily targets Chinese users. This article focuses on the officially verified capabilities listed in the introduction. Areas lacking installation commands and reproducible examples are honestly noted as not provided in official materials, with no speculation made. Key caveats include the restriction of the CNS literature scope and the need for manual verification of translation outputs.
Read MoreMDA: A Dedicated Skill Set for AI Agents
MDA is a dedicated skill set for AI Agents, with skills described in the universal SKILL.md format and compatible with multiple Agent platforms. It is community-published, versioned at v1.0.0, categorized under ai-agent, and tagged with requires_api_key=false, meaning no additional API key configuration is required. It addresses the problem of packaging the execution methods for MDA-related tasks into a standard Skill to avoid piecing together prompts each time. The materials provided do not include a specific list of sub-skills, trigger conditions, installation commands, or configuration fields, and this article does not engage in speculation. It is intended for developers whose tasks fall within the MDA scope, who use an Agent runtime that supports SKILL.md, and who prefer not to configure additional keys. It is worth noting that it is community-published, has zero installs, and lacks user feedback; therefore, its skill contents should be independently verified before deployment.
Read MoreAssistant for Prevention of Common MEP Installation Quality Defects: Organizing On-site Quality Issues into Closed-loop Rectification Plans
This is a Skill designed for on-site management personnel in mechanical and electrical (M&E) installation. It covers common quality defects across five major disciplines: water supply, drainage, and heating; electrical installation; cable tray fireproof sealing; lightning protection and grounding; and ventilation and air conditioning. By inputting five key pieces of information—the professional system, location, quality issue, construction phase, and objective—it generates a comprehensive report, including issue identification, matched defect type, cause analysis, key inspection focuses, corrective measures, re-inspection points, and a reminder for manual review. This article introduces the tool's coverage, output structure, installation method, typical usage scenarios, and defined boundaries.
Read MoreAgent Skill Creator: Adding the "Create Skill" Step to Agents
Agent Skill Creator is a community skill on SkillHub, identified by the coordinate @user_3c6cb52e/agent-skill-creator-sh-s6p2j4, version 1.0.0, categorized under ai-agent, sourced from the community, and requires no API key. It targets developers who need to extend AI Agent capabilities, transforming skill creation into a structured process rather than ad hoc prompt assembly. Due to the failure in fetching the official page, information such as the feature list, installation commands, and configuration fields were not included in the materials. This article only presents verifiable information: its purpose is to produce skill definitions rather than execute business tasks, and it must support a general SKILL.md format at runtime. Before using it, it is recommended to review the skill content and verify details after the official page is restored.
Read MoreCard Workshop (dsh-card-workbench): Orchestrating Multi-task Workflows in DeepSeek Harness with Cards
Card Workshop is a workflow plugin for DeepSeek Harness that breaks down requirements into 2–6 acceptable task cards, handling card dispatch, tracking, verification, and final review on the Host side. It can optionally take over existing work conversations and convert them into card groups. With no third-party runtime dependencies and local directory installation, configuration is written to .dsh-butler/settings.json. Based on the official README, this article introduces installation, task delegation and butler mode, permissions, and data paths.
Read MoreMr. Da Feng's Single-File HTML Packer (High Compression): Compressing multiple files into a "double-click-to-play" HTML
This article introduces the community Skill "大冯老师多文件打包单文件HTML(大幅压缩)" (Teacher Da Feng's Multi-File Packaging into Single HTML with Heavy Compression) on SkillHub, located at @indiv-dfteacher/fmg-htmldb, version 1.0.2, categorized under dev-programming, published by user_73cb5b59, sourced from the community. It addresses the problem of web-based mini-games, interactive courseware, and H5 works depending on external directories, requiring servers or decompression during distribution: it packages the entire project directory along with subfolders into a single offline-runnable HTML file, while heavily compressing images and audio to keep the output within a specified size budget (default 10MB), ultimately forming a "double-click to play" distribution format with a single file. The core capabilities are only three: multi-file merging, image and audio compression, and size budget control; HTML, JS, and CSS are not compressed—to further reduce size, code must be optimized independently first. The installation and activation section follows the standard SKILL.md format to explain the coordinates and page URL, does not fabricate installation commands, and notes that requires_api_key=false means no API Key is needed. The typical usage section only follows the three-step process given in the official introduction, clearly pointing out that parameters such as compression quality, format conversion, and over-budget strategies are not mentioned in the retrieved official materials and need to be verified in the SKILL.md or official page. Finally, it lists applicable scenarios and four limitations: it only handles distribution not development, benefits are limited when images are already compressed, over-budget behavior is unknown and it is recommended to test with small projects first, and with 0 installations and 929 downloads it is a relatively new community Skill with limited reference material.
Read MoreMarketing Customer Management: Delegating Customer Data and Campaigns to Skill
This article introduces the open-source Skill "Marketing Customer Management" on SkillHub (@org-02qudk26/salesforce-marketing-cloud-automation, Version 1.0.0, Source: enterprise, Publisher: Zhiyu Technology, Category: business-ops). Designed for marketing customer data and campaign management, this skill aims to reduce repetitive operations and directly generate actionable results. Its metadata indicates `requires_api_key=false`, meaning no additional API key configuration is required for activation. The article outlines the capability boundaries, available metadata, coordinates, official page URLs, and a three-step minimal verification path. However, because both official URLs were inaccessible during this verification process, no installation commands, directory paths, or configuration fields were provided, nor does it differentiate instructions for the Cursor, Codex CLI, and Claude Code environments. Furthermore, it notes the deployment status of version 1.0.0 (819 downloads and 0 installations) and clarifies that the rating is solely a platform tag and should not be used as a basis for selection.
Read MoreSkill Creation: Encapsulating Professional Expertise into Reusable AI Skills
Skill制作(@user_49033e54/skill-made, v1.0.0, community, ai-agent) is an AI Skill for the architectural design of AI Skills: based on user requirements, it encapsulates professional skills, workflows, or knowledge systems into reusable prompt-based skills, and produces a complete Skill plan following a standardized process that includes Role, Goal, Capability, Workflow, Output Format, and Rules. Typical trigger scenarios include creating or designing AI Skills, building AI assistants, turning domain expertise into Skills, and writing Skill prompts. It is not intended for creating native WorkBuddy platform SKILL.md Skills; for that, use skill-creator. Labeled as requires_api_key=false, it does not require an API key. The SkillHub page shows a rating of 100000, with 859 downloads and 0 installations; it is community-published and has zero installs, so verification is recommended before adoption. During this check, both the official page and API endpoint were unavailable. The provided materials do not include installation commands, directories, or configuration fields, and no unverified installation steps are described in the text.
Read MoreGeoMaster: Geospatial Science Skill Covering Remote Sensing and Spatial Analysis
Geomaster is a comprehensive geospatial science Skill released by Zhiyu Technology on SkillHub. Located at the coordinate @org-02qudk26/geomaster, it is classified as "professional" with version 1.0.0 and does not require an API Key. It integrates remote sensing (Sentinel, Landsat, MODIS, and SAR satellite image processing), GIS and spatial analysis, and machine learning for Earth observation into a single Skill. Covering over 30 scientific disciplines, it addresses the pain points of lengthy geospatial analysis workflows and fragmented tools. This article introduces its positioning, verified capabilities, activation methods, and applicable scenarios. It also notes that since official documentation lacks installation commands and reproducible examples, users should visit the official page to verify these details.
Read MoreArt Director Skill: Prompting the Model with Film Art Judgment Frameworks
This article introduces a community skill, "Production Design" (@user_3c6cb52e/self-ent-film-production-design, version 2.0.2, category: design-media, source: community, API key not required). Set in the context of film and theater production, it provides discussable judgment criteria centered on the production designer, covering three levels: camera language, narrative structure, and scene staging. This enables the model to go beyond generic visual descriptions and instead explain the narrative function, shot composition, and spatial flow of a scene. The trigger keyword is production-design, following the standard SKILL.md format. The article covers its positioning, core capabilities, installation and activation, typical usage approaches, and applicable scenarios and limitations. It also notes that because the official page is currently inaccessible, commands, configurations, and examples are not based on guesswork and should be verified against the official page once it is restored and scraped.
Read Moredsh-mini: Pop out DSH sessions as pinned WebView2 mini-windows
dsh-mini is a DSH client plugin maintained by xiaoxin3309 (npm dsh-mini 0.1.18, Windows x64 only). It pops out the current session into a native, always-on-top mini-window using WebView2, offering full-screen and selection screenshots, a "Ask Big Fat Fish" pill for text selection, and global hotkeys. To install, run `npm install dsh-mini` in `~/.dsh/profiles/desktop` and add the bundle name to `dsh.profile.bundles`. Requires Node.js ≥ 20, DSH ≥ 0.2.0-rc.1, and WebView2 Runtime. Based on the official README, this article describes the architecture, routing, configuration, and usage notes.
Read MorePDF & Image Table Extraction: Structuring Tables from PDFs, Scans, and Images into Excel
Introducing the community Skill "PDF Image Table Extraction" on SkillHub: Identify tables from PDFs, scans, images, and Office documents for structured export. It supports batch processing of multiple files, recognition of multi-row headers and restoration of merged cells, as well as merging tables across pages. The skill handles various data types—including dates, percentages, accounting negatives, leading zeros, and long numbers—and outputs Excel files with a table directory, alongside optional CSV and JSON formats. A TextIn API Key is required.
Read More