dsh-session-focus: Mark "Active" and "Suspended" Sessions in the DSH Sidebar
dsh-session-focus is a DSH workflow plugin maintained by Fan-sx (MIT, v0.1.2). It displays Focus (amber vertical bar) and Parked (hollow ring) markers before session titles, synchronized via clicks across three locations. When the workspace is collapsed, markers remain visible through folder dots and a "bar-in-hand" summary indicator at the bottom of the sidebar. Panel pages display entries grouped by useSessionStatus and custom markers. Purely client-side, with markers stored in localStorage and synced across tabs. Installation: dsh plugin --profile <profile> add dsh-session-focus.
Read MorePDF to Word: Convert PDF to Editable Word while Preserving Layout and Fonts
This article introduces the PDF to Word skill on SkillHub. It first detects the PDF text layer; text-based PDFs are converted directly with pdf2docx to preserve fonts, layout, and tables, while scanned documents are recognized via macOS Vision OCR and then structurally reconstructed into an editable Word file. No API key is required, making it suitable for contracts, scanned documents, judgments, and more. The article also explains its purpose, core capabilities, and trigger words, and frankly notes limitations such as: the official documentation does not provide installation commands, the scanned-document path relies on macOS, and layout restoration requires manual proofreading.
Read MoreSuper Versatile Macro-Micro Economist: One-Sentence Requirement-Triggered Integrated Macro-Micro Economic Analysis Engine
**Super All-around Macro-Micro Economist** (@user_8e3169aa/super-all-around-macro-micro-economist) is an economic analysis Skill based on the SKILL.md format. Responding to a single-sentence user request, it simultaneously invokes macroeconomic analysis frameworks (GDP accounting, growth, inflation, employment, monetary policy, fiscal policy, balance of payments, business cycles, and macro transmission mechanisms) and microeconomic analysis frameworks (supply and demand, elasticity, market structures, cost-benefit analysis, consumer behavior, enterprise decision-making, game theory/competition, and market failure). It then conducts an integrated analysis via a macro-micro linkage and transmission framework. The Skill integrates authoritative data sources—including the National Bureau of Statistics, the People's Bank of China, the General Administration of Customs, the Ministry of Finance, IMF/World Bank, FRED, and Wind/Choice—and employs a WebSearch/WebFetch cross-verification SOP to generate on-demand complete research reports, executive briefings, data visualization lists, and decision recommendation checklists. Originating from the community, it is classified as professional and requires no API Key.
Read Moredsh-matlab-feedback: Bringing MATLAB Desktop and Command Line into DSH Conversations in Real-time
dsh-matlab-feedback is a DSH client plugin maintained by Ezerruier (Windows, MIT). It captures read-only screenshots and command-line text from MATLAB and DPABI/DPARSF/SPM QC windows into the conversation via matlab_snapshot and matlab_console. The matlab_stage_code function can stage single lines of code to the command window for manual execution by the user. It requires Windows 10/11, PowerShell 5.1, MATLAB R2025a+, and DSH 0.2.1-rc series. Installation can be performed by pointing install_bundle in plugin_manager to the plugin directory, or by manually adding it to profile bundles.
Read MoreOverview of A-Share Income Statements: Transforming the "Profit Overview" Table from Choice Stock Diagnosis into a Self-Contained HTML Report
The A-share Income Statement Overview is a community Skill on SkillHub (ID: @user_058048a4/profit-overview, v1.1.0, category: professional, requires_api_key=false). Upon inputting a stock code or name (e.g., 688018.SH, 300308, Kweichow Moutai), it automatically retrieves income statement, balance sheet, and cash flow data from Eastmoney. It calculates metrics such as revenue, net profit attributable to the parent, core profit and growth, contribution rate of core profit changes, profit bridge analysis, profitability (gross margin/net margin/core profit margin/ROE/capital returns), turnover days, and cash flow quality. It generates a self-contained HTML report with inline ECharts, featuring 16 charts rendered strictly according to original Excel specifications, displayed within the WorkBuddy preview panel. Trigger words include "Profit Overview," "Income Statement Overview," "Core Profit Analysis," "Income Statement Analysis," "Profitability Analysis," and "Analyze this company's profits." This article details its positioning, core features, activation methods, typical use cases, and usage considerations.
Read Moredsh-qq-shell-commands: Execute Shell Commands on DSH Host Machine via QQ Private Messages
The DSH Host plugin dsh-qq-shell-commands receives private QQ messages via NapCat (OneBot 11) over WebSocket. It matches them precisely against configured trigger phrases and executes the corresponding shell commands in the host machine’s current user environment (not within the DSH sandbox). The plugin then sends the execution result back to the user through direct messages, utilizing the prerequisite plugin dsh-napcat-tools. This plugin is suitable for automated recurring tasks and hosting scripts, particularly in scenarios where users prefer not to grant execution permissions to the DSH sandbox. Based on the official README, this article outlines the prerequisites, installation commands, configuration of qq-shell-commands.json, and typical usage patterns. It also highlights the associated Remote Code Execution (RCE) risks and emphasizes defining appropriate security boundaries.
Read MoreNew Stock Research Report Skill: Standardize the IPO Initial Coverage Analysis Process
This is an in-depth research Skill for new stocks (IPOs) based on the Hundsun Juyuan MCP financial database, located at @user_cc9425b3/jy-ipo-research. Version 1.0.2, source: community, category: professional, with the tag requires_api_key=true. It covers core chapters such as company fundamentals, industry position, valuation analysis, fundraising and investment projects, investment outlook, and risk factors, and produces an investment analysis report with a first-time coverage rating. Trigger scenarios include pre-subscription analysis, first-day analysis, and in-depth research reports, covering newly listed companies as well as prospective issuers that have passed the review meeting, been accepted, or been subject to inquiry. The article explains its positioning, verified scope of functions, and trigger usage, and points out that during this fetch the official page could not be accessed, and the installation commands and the main body of SKILL.md were not obtained. Therefore, it does not fabricate configurations, and reminds users to prepare data credentials and run through a real new stock once to verify standards.
Read MoreComprehensive Numerology Deduction Assistant: A Structured Interpretation Skill Triggered Only Upon Explicit Requests for Numerology Deduction
The Comprehensive Numerology Deduction Assistant is maintained by community publisher user_ddfdfa23 (Version 0.1.1, Category: life-service). Its description states that it is designed to synthesize Eastern and Western numerology, symbolic analysis, and modern narratives to generate structured interpretations when explicit numerology deduction requests are made, thereby preventing erroneous triggers during general psychological analysis. This article verifies its capability boundaries, metadata, and official address, explains the reasons for the failure to retrieve the SKILL.md content and installation commands, and provides an integration workflow based on the standard SKILL.md format, along with recommendations for positive and negative trigger tests.
Read MoreRegex Pattern Generator: Integrating Common Patterns and Real-time Validation into a Single Skill
The Regular Expression Generator is a developer-oriented tool for generating and validating regular expressions, released by Changsha Xingwang Intelligent Technology Co., Ltd. It is classified as dev-programming and is version 1.0.2. The tool features 19 built-in common patterns (including phone numbers, email addresses, ID cards, URLs, dates, IP addresses, and currency amounts) that can be generated with one click. It supports custom rules for real-time matching tests and interprets the meaning of existing regular expressions. This tool is suitable for developers who need to validate data formats, extract or replace text fields, or review regular expressions written by others. The article outlines its positioning, core features, usage paths, applicable scenarios, and limitations, noting that the official documentation does not provide installation commands, which should be based on the official page.
Read Moredsh-qq-approval-notify: QQ Notifications for DSH Approvals and Inquiries
The Host-side DSH workflow plugin maintained by EviMing monitors approval/request and user-questions/request events. It formats interface prompts, such as permission requests and ask_user_question queries, into QQ private messages and sends them to the ownerQQ via the local NapCat OneBot. In cases where the service is unavailable or sending fails, it falls back to Windows Toast and tray balloon notifications; if these also fail, the system abandons the notification without affecting the approval process. Configuration is done via qq-approval-notify.json, and installation via the plugin_manager using the tgz package (generated via `npm pack`) is recommended. This plugin is for notification purposes only; it does not perform approvals or answer questions on the user's behalf.
Read More[Zhanshen Digital Tech] Copywriting Skills to De-AI-ify Content: Adding a Round of Human Evaluation After Rewriting
A skill designed to remove AI traces from copywriting for Xiaohongshu, Video Accounts, Official Accounts, and Douyin. Produced by Zhanshen Digital Technology, it is located on SkillHub at @org-tj8dikyv/de-ai-humanizer. Classified under content-creation with an enterprise source type, the official entry (Version 1.0.0) notes that no API key is required. The core process involves two steps: first, rewriting the copy to resemble human writing by eliminating AI traces and lowering AI detection rates; second, simulating readers with distinct user personas to evaluate the text paragraph by paragraph. This step identifies "what still sounds like AI" versus "what doesn't sound human," providing a "humanness" score and pinpointing residual AI traces. The key difference from simple rewriting tools is the inclusion of a verification step from a reader's perspective after rewriting. The official description states, "While others only rewrite without verification, this product allows real people to evaluate and gatekeep after the revision." Official trigger words include: copy de-AI, social media de-AI, remove AI flavor, lower AI rate, AI trace removal, pass AI detection, humanize rewriting, real-person evaluation, reader acceptance, real-person trial reading, and AI trace detection. As the official page scraping failed during writing, the installation command and reproducible examples could not be verified; consequently, no commands or configurations were fabricated in the text. Furthermore, as the ratings and download data were inconsistent across different sources (100000/1017 vs. 99788/943), this has been honestly noted, with the official page serving as the source of truth.
Read Moredsh-approval-timeout: Adding timeouts for DSH authorization and question waits
DSH blocks while waiting for user authorization or responses to questions. The community plugin `dsh-approval-timeout` (EviMing, admin-security) introduces configurable deadlines for `approval` and `ask_user_question` requests in the Web client (default: 10 minutes). Upon timeout, it concludes the process as rejected or skipped, extending the timeframe by 60 seconds if the user is currently editing. It does not perform auto-approval or handle plan reviews. Settings are persisted in `localStorage`, with installation commands available in the GitHub README. This has been tested on `dsh 0.2.0-rc.2`.
Read MoreMARS - The First Large Model for Pharmaceutical Marketing, Sales, and Medical Strategy
MARS is a vertical large model for business diagnosis in the pharmaceutical industry, published on SkillHub by user_557cd010. The current version is 4.2.3, categorized as business-ops, and sourced from the community. The official page简介 states that it has 22 built-in core tools covering five major capability domains: brand strategy, market access pathways, medical support, multi-dimensional growth, and regional management. The tag requires_api_key=false is set, with a page score of 100000, 1009 downloads, and 0 installations. This article identifies conflicting data: scraped clues mention eight capability domains and 19 tools plus a "Tool 22 routing engine," which contradicts the official page简介. Furthermore, claims regarding the integration of methodologies from McKinsey, BCG, Bain, ZS, IQVIA, PROVEN, Hua & Hua, and Jinqian Uncle could not be cross-verified on accessible pages. Due to a failed scrape of the official page, no installation commands, configuration fields, or usage examples are provided; users are advised to verify details independently before integration based on the coordinates @user_557cd010/mars-pharma-expert. The article also outlines applicable scenarios and limitations: while suitable for structured tool support across various stages of pharmaceutical marketing, given the zero installation count and page instability, all outputs require review by business professionals.
Read MoreEfficient Multi-Agent Teams for Data Analysis: 5 Agents Covering the Entire Data Processing Pipeline
This article introduces the "High-Efficiency Data Analysis Multi-Agent Team" Skill available on SkillHub. Released by user_d76c672d (Version 1.0.5) from the community and categorized under "data-analysis," this solution utilizes five specialized agents to sequentially execute tasks: data collection, cleaning, statistical analysis, visualization, and report generation. It supports various formats, including Excel (xlsx/xls), WPS Spreadsheets (et/ett), CSV, and TSV. The tool automatically recognizes Chinese headers and handles currency symbols (¥) and thousand-separator commas within monetary values, outputting charts in both SVG and PNG formats. Compatible with conversational triggers, direct script execution, and a Python API, the developers state it features a self-improvement mechanism and supports offline operation within China. While the page indicates a rating of 100000, 978 downloads, and 0 installs, with no API key required (`requires_api_key: false`), the retrieved materials do not provide specific installation commands or configuration fields. The article honestly addresses this absence while providing the official source URL.
Read Moredsh-screenshot-tool: Capturing Windows Desktop in DeepSeek Harness
dsh-screenshot-tool is a DSH client plugin maintained by Eva7788519. On Windows, it captures PNG images of the virtual desktop using the screenshot tool and the Composer camera button. The tool saves images to the session directory, where they can be read back using `read_image`. The button supports preview, download, and clipboard functions. To install, run `dsh plugin --profile <profile> add dsh-screenshot-tool`; a Harness restart is required after installation. This tool is Windows-only and relies on PowerShell and GDI for full-screen capture.
Read MoreCampus Psychological Crisis Intervention Assistant: Transforming Crisis Assessment and Intervention Processes into Reusable Skills
The Campus Psychological Crisis Intervention Assistant is designed for student crisis events. It provides crisis assessment and grading, generates intervention workflows based on the six-step method, offers communication scripts for home-school communication with four types of parents, and provides templates for record-keeping and continuous tracking. It is suitable for psychological counselors and homeroom teachers to quickly organize responses when signs such as self-harm, suicidal tendencies, or psychological crisis warnings arise. It is not intended for routine psychological counseling, lesson plan design, or psychological scale development. During use, safety must be prioritized, and high-risk situations must be referred to professional medical institutions.
Read Moredsh-codebuddy-acp: Integrating Local CodeBuddy CLI into DeepSeek Harness
dsh-codebuddy-acp (v0.5.0, EVA09) registers the local CodeBuddy Code CLI as a CodeBuddy provider for DSH via the ACP protocol. Upon installation, the model selector displays a CodeBuddy group featuring credit multipliers and configurable thinking levels, while the sidebar offers a console for credit and model management. Tool calls are replayed as DSH tool cards via tool-call blocks and `codebuddy_tool`, synchronizing titles, todos, and approvals. To install, use `dsh plugin add` with a local path or `install_bundle` as described in the README. Users must be logged into CodeBuddy CLI prior to use. Please note that the plugin operates with the same permissions as DSH; review the source code before deploying.
Read MoreOne Template, Ready Documents: Applying Institutional Word Templates to Notices, Schemes, and Lesson Plans
This article introduces the community Skill “Kangkang Yiyang · Chengzhang” (handle: @user_5b5e3b96/yiyang-chengzhang, version 0.3.0, category: office-efficiency, publisher: user_5b5e3b96). The official description states that it can initialize or reuse an organization’s Word template and, based on the confirmed font and layout, generate documents one by one or in batches—including notices, plans, summaries, records, and lesson plans. The article verifies the information shown on the public page: the tag requires_api_key=false, with a page score of 100000, 1060 downloads, and 0 installs. Because no unified installation command, differences across Cursor/Codex CLI/Claude Code, configuration fields, and details on resuming from breakpoints are publicly provided, the article only describes verifiable capabilities and does not include speculative commands. A typical workflow is to first confirm the template layout, then run a single trial before batch producing similar documents. It is suitable for scenarios where a template already exists and there is a clear need for batch document generation; the preservation of effects for complex Word features must be tested in practice.
Read MoreAnatomy Study: A Guide to Human Anatomy Drawing for Painting
Anatomy Study is a community-sourced design-media skill maintained by user_3c6cb52e, with the current version 2.0.2 and no API key required. It serves as a guide to drawing the human anatomy, covering skeletal structure, muscle groups, and dynamic human figures, intended for use in scenarios such as character design, illustration, and comic panel layouts. It enables the model to describe the human body by breaking it down to the anatomical level, referencing specific bones and muscles. The materials do not provide the full SKILL.md text, installation commands, or configuration fields. This article only presents verified information. The installation entry points to the official page; it is recommended to validate output quality with a small sample before installation.
Read MoreSupply Chain Management Skill: Methodology and Practical Framework for Entrepreneurs
Introduction to the positioning and scope of the Community Edition Skill "Supply Chain Management": Published by user_3c6cb52e; source: community; category: business-ops; version: 2.0.2; no API key required. Official description: "A practical guide for business and entrepreneurship: Supply Chain Management—systematic methodology and an operational framework, helping entrepreneurs go from 0 to 1." It is suitable for startup teams without an established supply chain structure as a framework for itemized discussion. Community data: rating 100000, downloads 1006, installs 0, for reference only. Official materials do not provide the installation command or client integration differences; no speculation is made here. Please refer to the SkillHub page and SKILL.md.
Read MoreDialect Master: Agent-oriented Bidirectional Mandarin-Dialect Translation Skill
Fangyan is a community-sourced Agent Skill categorized as "education" (version 1.0.0, publisher: user_bf6d22bb, requires_api_key=false). It is triggered when users request conversations in a dialect, ask to translate content into a dialect, or mention specific Chinese dialects such as Northeastern, Sichuanese, Henanese, Cantonese, Shanghainese, Hunanese, or Shandongese. Its core capability is to convert Mandarin into authentic dialectal expressions and converse with users in a specific dialect style. The article is structured as follows: Preface, What is this, Core Functions, Installation and Activation, Typical Usage, Use Cases and Notes, and Conclusion. It should be noted that the retrieved materials do not contain installation commands, installation directories, or configuration fields. As both the SkillHub page and the official API address failed to fetch during this session, this article refrains from providing unverified installation commands. It only confirms that no API Key is required and suggests referring to the official channel address for authoritative information. Usage examples are categorized into three types based on the official introduction's trigger conditions: dialect conversation, rewriting Mandarin into dialect, and content adaptation by region. The article also cautions that the authenticity of the dialect relies on corpus coverage, multi-turn dialogues should lock onto the same dialect, and formal documents are not suitable for direct dialect rewriting. The data shows a rating of 100,000, 994 downloads, and 0 installations; the zero installation count indicates a lack of public records regarding large-scale deployment.
Read MoreCommercial-Ready Local LLM Deployment: Clarify "What to Install and Desired Outcomes" Before Executing the Command
This Skill targets developers aiming to run open-source LLMs on their own machines or LANs, solving the issue of disjointed components—where inference engines, web clients, and agent frameworks are installed separately with incompatible APIs. It covers inference engines such as llama.cpp, Ollama, LM Studio, and vLLM; clients like OpenWebUI, Cherry Studio, and LobeChat; as well as frameworks including Dify, FastGPT, Flowise, Crawl4AI, and Hermes function calling, unifying them all via OpenAI-compatible APIs. Regarding workflow, the Skill first confirms "what to download" and "what effects to achieve" before starting. It then completes environment diagnostics, downloading, startup, and connectivity verification without interruption, reporting progress every 5 minutes during the download phase. Note before enabling: The Skill itself does not require an API Key, but model weights and inference engines must be prepared by the user; it is responsible only for deployment and connectivity, not for training or fine-tuning. Official URL: https://api.skillhub.cn/user_cb80f343/local-model-deploy, Coordinates: @user_cb80f343/local-model-deploy, Version: 1.0.0, Source: community, Category: it-ops-security. It is recommended to perform a minimal validation test before adoption.
Read MoreSkill for AI Video Production for Finance & Tax Lead Generation: Converting Licensed Materials into Publishable Short Videos
This is a "Finance & Tax Bookkeeping AI Customer Acquisition Video Creation" Agent Skill published on SkillHub. It accepts authorized reference images of individuals, business environment photos, and product or service materials. It covers industry customer acquisition strategies, AI avatar consistency, creative direction, storyboarding, voiceovers, audio, and conversion actions, while generating video prompts optimized for Jimeng and Seedance. It is suitable for finance and tax bookkeeping stores, institutions, brands, or service providers dealing with industry promotional videos, AI replacing live-action appearances, short video customer acquisition, video generation from reference images, video prompt engineering, and material rework. Before use, please confirm material authorization, reference image quality, voiceover compliance, and the generation results on the target platforms.
Read Moredsh-read-image-compress: Compress images before read_image enters the session
Introducing the `dsh-read-image-compress` plugin for DSH, maintained by etherfun. This plugin allows for independently toggling resolution and bitrate compression during the `tools/post-execute` phase of `read_image`. It utilizes `sharp` to submit smaller attachments while replacing the tool's content, and displays the original image, compression ratio, and estimated DeepSeek tokens at the bottom of the card. This document outlines the web profile installation command, the Plugins page, and the `cordis.patch.yml` configuration, as well as limitations such as Sharp fallback and its exclusive application to newly read images.
Read Moredsh-starsector-mod-workflow: Running the Starsector Mod Development Workflow in DSH
dsh-starsector-mod-workflow, maintained by erenamochizuki, is a port of the Codex skill starsector-mod-workflow to a DeepSeek Harness plugin. A single package registers both the skill (including resources like REF/KIT/API/Script and DSH environment instructions) and the starsector_moci tool, and binds to Moci projects on Windows via moci.json, supporting CLI commands such as doctor, compile, build, verify, sync, api-src, and tutorial. Requires DSH >= 0.2.0-rc.2 and Node >= 18. It can be installed via the plugin marketplace custom source github:erenamochizuki/dsh-starsector-mod-workflow or by using `dsh plugin add`. A restart is required for it to take effect. Released under the MIT license, version 1.0.0.
Read More