Nico-WorkBuddy-Deskpet: A State-Aware Desktop Pet for WorkBuddy
Nico-WorkBuddy-Deskpet is a community skill that transforms WorkBuddy into a desktop pet permanently residing at the bottom-right corner of your screen. It is fully compatible with Mac, Windows, and Linux. The pet detects four distinct states—thinking, working, completed, and failed—and dynamically updates its appearance accordingly. Users can perform one-click avatar swaps using either AI-generated or self-provided chibi images, with no API key required. Installation is seamless: simply type "install the desktop pet for me" in the chat. This article outlines its positioning, core features, installation and activation steps, typical use cases, and important usage notes, while also providing the official link.
Read More**High EQ Response & Relationship Decoder: Translating Conversational Tone and Relationship Signals into Ready-to-Send Replies**
Reply-like-a-pro is a community-sourced Agent Skill located at @user_6935a801/reply-like-a-pro (Version 1.0.0). Classified under content-creation and published by user_6935a801, it requires no API Key. Simply paste text or a screenshot from WeChat, QQ, DingTalk, SMS, or direct messages, and the skill will first analyze the tone, subtext, relationship signals, and communication risks. It then provides three natural, non-cheesy, high-EQ replies ready to send instantly. It covers scenarios including romantic ambiguity, workplace communication, and client follow-ups, as well as rewriting for refusals, apologies, consolation, invitations, reminders, and boundary setting. The documentation also outlines capability boundaries, usage instructions for the Agent side, and precautions, while providing the official URL and SkillHub page.
Read MoreNiu Lai Style Generator: Applying the Artistic Style of Niu Lai to Images and Short Videos
The Niulai Style Generator (@user_ffe5c7f8/niulai-style) is a style generation Skill built on the visual characteristics of the 2026 animated film *Niulai*. It was published by community developer user_ffe5c7f8, falls under the design-media category, and is currently at version 1.3.0. It offers two artistic styles: a Chinese-style ink wash poster aesthetic and the low-precision 3D animation meme style actually used in the film, while supporting both image and short-video outputs. The Skill includes built-in Chinese-English bilingual prompt templates and scene classifications. Once the style, output format, and scene are selected, generation can be triggered in one click, with no additional style parameters required from the caller. No command-line installation commands, directory paths, or configuration fields are provided in the documentation. Although the Skill is marked as requires_api_key=false, the specific drawing backend it depends on and who provides the quota are not specified, so users should verify this on the official page before enabling it. The article also outlines a typical usage workflow and draws attention to considerations such as the fact that style fidelity relies on built-in templates, short-video specifications are unknown, and the maintainer is a community developer.
Read MoreNutrition Log of Chinese Cuisine: Converting Home-style Dishes and Delivery Meals into Calories and Protein
This is a community-published Skill at @user_36c7efa2/nutrition-log-cn, version 1.0.0, categorized under life-service. It addresses the lack of a Chinese food calorie database by breaking down home-style dishes, takeout, and cafeteria meals into estimable components, providing rough calorie, protein, and dietary-structure logging. Core modules include meal component breakdown, calorie estimation, macronutrients, takeout substitutions, and logging templates; trigger keywords cover diet logging, calorie estimation, takeout, nutrition, and fat-loss meals, with no API key required. It estimates calories and macronutrients for common Chinese dishes by decomposing them into components when precise data is unavailable, and records overall dietary patterns. Use it when logging meals, evaluating takeout, planning fat-loss diets, or tracking protein and macronutrient intake. This article describes its positioning, capabilities, typical usage, limitations of rough estimates, and includes the official URL.
Read Moredsh-lit-library: DSH Literature Library with Budget Indexing and Verbatim Checking
dsh-lit-library is a DSH memory plugin maintained by hellocyhgy, designed to manage papers by project area: lit_import deduplicates entries and extracts full text (with one index line per reference), while lit_check verifies budgets, broken links, and literal consistency of excerpts. It is not a PDF reader or search tool; users must install dsh-lit-library via bundles or install_bundle in their profile and restart (requiring DSH >=0.2.0-rc.2). Optional configuration is available for scanned document OCR. Released under the MIT license.
Read Moreqmt-trader: Unified CLI Driver for the Great QMT XunTou Quantitative Trading Terminal
`qmt-trader` is a community-sourced Agent Skill (coordinate: @user_70fc5396/qmt-trader, version 1.0.1, category: professional, tag: requires_api_key=false) that encapsulates the capabilities of the QMT XunTou quantitative trading terminal into a unified CLI script. Functionalities cover real-time market quotes, historical K-line data, account assets and holdings queries, order and transaction queries, buy/sell order placement, and order cancellation, as well as sectors, Dragon and Tiger list data, northbound capital, and financial data. It also includes installation and deployment guidance for the built-in `xtquant_big_convert` bridge service, covering package installation, QMT terminal file synchronization, configuration, startup verification, and troubleshooting. It is suitable for scenarios such as LLM-assisted quantitative trading analysis, market trend assessment, position monitoring, and semi-automated order placement. The retrieved materials do not provide copy-paste ready installation or example commands; this article provides a calling framework based on the capability list, but specific details should refer to the SKILL.md file within the skill package. Order placement operations connect directly to the broker's counter and involve real capital operations, so verification is required before use.
Read MoreCriminal Defense Full-Process Intelligent Workbench: Case Collaboration Skills for Ten Workflows
A Community Skill designed for criminal defense lawyers in Mainland China, released by user_dd647bf8 at @user_dd647bf8/cdiw-core, version 2.2.0, categorized as professional. It decomposes the defense workflow of a single case into ten operational lines: client meetings, document drafting, deadline management, family liaison, agency communication, compulsory measures, investigation and evidence collection, legal research, sentencing circumstances, and procedural compliance. Operating in tandem with a dual-mainline structure of ten skill libraries, it covers the entire scope from case file creation to procedural compliance. The Skill maintains case status and service progress files, supports resumable workflows and cross-line coordination, and features built-in statutory validity checks, deadline calculations, sentencing estimates, and automatic data masking. All 20 scripts run locally with zero network requests, generating a master working draft for the responsible lawyer's review. This version excludes original legal text packages, downgrading statutory verification to a "to be verified" status. It is not applicable to civil, administrative, or non-litigation matters, supervisory investigation stages, cases outside Mainland China jurisdiction, or non-legal requests, nor is it intended to replace legal judgment, promise outcomes, forge evidence, or circumvent laws.
Read Morefupan-assistant: A guided reflection assistant that turns "can't-tell-good-from-bad" feedback into reusable experience
fupan-assistant is a review-guide Skill published by community developer user_d7fa4d48, categorized under knowledge-management with version 1.0.0. Rather than writing the review document for the user, it asks probing questions across five progressive levels—ranging from "how did it feel" to "what actually happened", "what are the root causes", "can a pattern be extracted", and "what core competencies can be consolidated"—ultimately generating a review card for reference when handling similar tasks in the future. Typical use cases include project reports, client visits, and requirements requiring iterative revisions with AI. To use it, simply issue a review request in natural language, with no API Key configuration required. Currently, on SkillHub, it has a rating of 100000, along with 933 downloads and 0 installations.
Read Moredsh-git-tab: Review Git changes and commit with one click in the DSH Web sidebar
dsh-git-tab (hejingzhou/dsh-git-change) is a DeepSeek Harness Web plugin that provides a "Git Changes" tab in the right sidebar of the session. It lists staged, unstaged, and untracked files with line statistics, unified diffs, and a git log topological commit tree. It also features one-click local or remote commits, ff-only pulls, and branch switching in a clean workspace. The installation command is `dsh plugin --profile web add github:hejingzhou/dsh-git-change` or the npm package `dsh-git-tab`. The plugin treats the session cwd as the repository root and executes git using the dsh process permissions. Licensed under MIT 1.0.0.
Read MoreTrademark Application Assistance: Category Planning and Material Preparation for China Trademark Applications Skill
The Trademark Application Assistant (@user_55833ee9/trademark-assistant, v1.7.2, community-sourced, categorized under business-ops) is a Skill designed for Chinese trademark applications. Based on the Nice Classification and referencing Chinese laws and regulations, it provides three capabilities: category planning, preliminary screening of registrability, and application material preparation, outputting structured recommendations and risk grading. No API key is required. The official documentation explicitly states that it does not replace formal legal advice, does not guarantee registration success rates, and does not handle complex trademark dispute cases. The article introduces its positioning, core features, enablement methods, typical usage, and applicable boundaries.
Read MoreWhale Whispers (dsh-whale-whispers): Quote Tab in DSH Session Header
Whale Whispers, a DSH client plugin maintained by heiyedehen, provides a "Whale Whispers" tab in the conversation header, enabling agents to atomically write Markdown quotes via the `whale_whisper_append`, `list`, and `remove` tools. It supports contextual annotations, file renaming and path changing, as well as exporting anonymous versions. The installation command is `dsh plugin add dsh-whale-whispers`, requiring a Web client; the target file can be specified via `quotes.config.json` or environment variables. The panel only reads the file when switched to and does not automatically record every line of conversation.
Read MoreMeeting Decision Amplifier: Transforming Vague Agenda Items into Verifiable Decision Frameworks Before Meetings
Meeting Decision Amplifier is an "office-efficiency" Skill published in the community, located at @user_00c9b356/meeting-decision-amplifier, currently at version 1.3.2. Instead of simply filling in agenda templates, it deconstructs vague agenda items into four components before the meeting: defining the real problem, listing the evidence required for decision-making, setting decision criteria, and identifying boundaries where conclusions become invalid due to changing conditions; the latter borrows from Munger's approach to conditional variables. The Skill is adapted from the Cognitive Amplifier v5.1 three-step loop engine and integrates with Tencent Meeting Skill to fetch historical meeting data for participant profiling. It is triggered via natural language, such as "Help me prepare for a meeting," "How should I run this meeting," or "Analyze with the decision amplifier." Note: The official page and API address could not be accessed this time; therefore, installation commands and configuration fields cannot be verified. Only the page address is provided in the text for manual verification. The tag requires_api_key=false indicates only that the Skill itself does not require an additional API key, and does not imply that Tencent Meeting capabilities are independent of credentials. The page shows a rating of 100000, 879 downloads, and 0 installs. As this is a very new community Skill, it is recommended to pilot it on a small scale first.
Read MoreAI Hallucination Checker: Tracing Each Citation in Reports
The AI Hallucination Checker, released by Bowei Management Consulting, is classified under knowledge-management with version 1.0.0, and performs item-by-item verification of citations in reports, PPTs, Word documents, and PDFs. It classifies citation status into five categories, such as traceable-to-source, fabricated-cobbled-together, and unverifiable, and outputs a red-flag list by severity. The methodology is labeled as the GPTZero five-stage pipeline. The currently retrieved materials do not provide verifiable installation commands or configuration fields; before actual integration, refer to the official SkillHub page or SKILL.md as authoritative. It is suitable for citation risk screening before document delivery, but unverifiable does not necessarily mean incorrect; manual source tracing is still required.
Read Moredsh-openspec-pipeline: Organizing OpenSpec Changes into a Pipeline and Executing According to the Diagram
The DSH workflow plugin `dsh-openspec-pipeline`, maintained by heibaoxia, uses `openspec/pipeline.json` to arrange multiple changes into a dependency graph, enabling sequential or parallel top-level conversation execution within the panel workspace. Task and planning progress are read from `openspec list` and `status` respectively, with changes being marked as complete only when they enter the archive. When a change stalls, users can run `validate --strict` and then manually archive it. Based on the GitHub README and package.json, this article covers installation commands, panel usage, permissions, and considerations. Current version: v0.19.0, MIT license.
Read MoreCAJ to PDF Conversion Skill: Converting CNKI CAJ Files to Readable PDFs
This article introduces a CAJ to PDF conversion skill designed to transform CNKI CAJ-format documents into PDF files. It supports both individual file conversion and batch processing of entire directories, built on the open-source caj2pdf tool. The article outlines the skill's positioning, core functionalities, the underlying tool's actual capabilities, and its known limitations. It also provides reproducible command examples for caj2pdf, making it a useful reference for developers needing to process CNKI CAJ documents within Agent workflows.
Read MoreWeb Automated Check-in: Integrating a Universal Browser Check-in Engine with Agents
Web Auto Sign-in is a community Skill posted by user_8f6c5b09 (@user_8f6c5b09/web-signer, v1.4.0, category: ai-agent). It is designed as a general-purpose browser auto sign-in engine and requires an API key. It targets sites that involve human verification, Cloudflare protection, or QR-code login, where an Agent drives the browser to complete login and sign-in. Supported sites are concentrated in the scripts/sites/ directory, and new sites are integrated via configuration rather than rewriting scripts. It covers three trigger scenarios: onboarding a new site; keyword-based triggering (e.g., website sign-in / auto sign-in / add a sign-in site); and performing sign-in on already-supported sites, as well as re-login and status queries. This article explains its positioning, core capabilities, and usage, while also noting considerations such as community provenance, credential requirements, and the possibility of page/layout changes over time.
Read MoreBilibili Video Notes Generation and Knowledge Organization: Transforming Public Videos into Structured Learning Notes
This article introduces the Skill "Bilibili Video Note Generation and Knowledge Organization" (Version 1.0.2, categorized under education) published by community author user_113322ce. By inputting a public Bilibili video link, the Skill automatically generates six types of structured content: learning objectives, knowledge outlines, core concepts, case methods, review questions, and action checklists. It is triggered when users request note organization, knowledge extraction, outline generation, or review checklist creation. The Skill explicitly does not cover entertainment content marketing analysis, verbatim subtitle export, or interaction data queries. When a link is inaccessible or subtitles are missing, it reports the situation honestly without fabricating content. No API Key is required for use. This article explains its capability boundaries, typical triggering use cases, target audiences, and limitations, and provides the official address.
Read Moredsh-session-merge: Merging and Session Appending Using the DSH Write API
The DSH desktop plugin dsh-session-merge (v0.6.0, MIT), maintained by HCQM, provides categorized memory. It resolves the difficulty of manually assembling multi-session history files: operations such as fork cloning, fullCopy/mixCopy, and fromJsonl all utilize the DSH write capability via session_merge_build calls, implementing three-layer validation and planCompact cut points. Install via `dsh plugin --profile desktop add dsh-session-merge`. Tested for compatibility with DSH 0.2.1-rc.5 and Session v4 formats; for large-scale writes, it is necessary to restart and resume after autoCompact.
Read MoreFliggy: Zero-Config Travel Search Skill
Fliggy Travel is a lifestyle-service Skill published on SkillHub. Version 1.0.3, it is maintained by community developer user_b95ee7e5. It encapsulates travel searches—hotels, flights, attraction tickets, train tickets, and so forth—into nine tools, supports natural-language queries, and returns results that include booking links and refund/change policies, with data originating directly from Fliggy’s official feed. The Skill declares requires_api_key=false, so no API key or similar credentials are needed; queries can be initiated immediately after installation, making it suitable for developers who want a quick travel-search entry inside an Agent without going through the platform’s API onboarding process. The article outlines its core capabilities, installation and activation steps, natural-language usage examples, and applicable scenarios. It also highlights caveats, including stability risks associated with community maintenance, the conversion-oriented nature of external links in the results, and the fact that it serves purely as a query entry point rather than a full itinerary-planning engine.
Read MoreSkill for De-AI Human-Like Writing Paradigm v6.1
Published by user_7e6f1fc8 on the SkillHub community, this Skill is categorized under content-creation (Version 6.1.0) and requires no API key. It is driven by a dual mechanism of code and prompts. The `detect.py` script first performs style routing, followed by quantitative detection of AI fingerprints across 18 dimensions (12 for fiction and 6 for non-fiction). It features built-in domain terminology protection, scenario direction flipping, automatic style discrimination, and style-specific red lines. The six non-fiction dimensions are integrated from the 24 modes of `humanizer-zh`. The prompt layer provides the Nine Laws, Four Paradigms, Eighteen Filters, a revision ladder, and an example library. Trigger words such as `/去AI味`, `/过检`, `/仿人写作`, `/AI味检测`, `/降AI率`, and `/仿真人写作` cover both detection and rewriting actions. Officially, it has been regression-validated at 23/23 using the v4.2 manual and actual readings from `humanizer-zh`. It is applicable to novels, non-fiction, current commentary, and WeChat Official Account scenarios. Limitations include that the dimensions only measure tone and do not evaluate content quality, and articles straddling style boundaries require manual confirmation first.
Read MoreOfficial Document Writing Standards Assistant: Integrating GB/T 9704 Format Specifications into the Writing Workflow
The Official Document Writing Assistant is a community-sourced Skill located at @user_af163eaa/gongwen-format-assistant (v1.0.0), categorized under office-efficiency. It provides document structure (header, body, and footer), templates for six common document types (requests for instructions, reports, notices, letters, meeting minutes, and opinions), standard terminology, and a formatting checklist based on the "Format for Official Documents of Party and Government Organs" (GB/T 9704). It activates when the user requests or mentions writing official documents, requests for instructions, reports, notices, or red-header documents, and serves exclusively as a reference for format specifications. The tag requires_api_key=false indicates that no API Key is required. The article introduces its positioning, core features, official address, enablement methods, and trigger-based usage, while clarifying usage restrictions such as acting solely as a formatting reference rather than drafting the content.
Read Moredsh-rea-dispatcher: Consolidates Over 200 REA Reverse Engineering Tools into Three DSH Commands
dsh-rea-dispatcher is a DSH client plugin maintained by HarunaTyan. It folds the 207 reverse engineering MCP tools provided by rea-agents into rea_tools, rea_help, and rea_call, with optional six read-only domain entry points, significantly reducing context usage compared to directly mounting REA. Before use, you need to prepare Node.js 22/24, rea-agents, and Ghidra 12.1.x yourself. The article explains the installation command, cordis.patch.yml configuration, typical AI invocation chains, as well as boundaries such as cwdAllow and sessions not persisting across rounds.
Read MoreProduct Manager: One-Sentence Full-Auto Product Creation (Requirements-Design-PRD) — An Agent Skill for Propelling a One-Line Requirement into a Deliverable PRD
This article introduces the SkillHub community skill "Product Manager: Create Products Fully Automatically from a Single Sentence (Requirements–Design–PRD)", located at @user_8c7f0f65/pm888 (version 4.2.1). It is categorized under ai-agent and published by user_8c7f0f65. Starting from a one-sentence requirement, it walks through requirement decomposition, solution design, and PRD generation, and includes an adversarial review mechanism that challenges the user's plan with a 0–40 scoring matrix (a 0–40 scoring rubric), uses the watermelon defense (an approach that exposes progress illusions), and applies decision pre-registration to counter post-hoc rationalization. It requires no API key (requires_api_key=false). The article describes the installation URL and the showcase page, and notes inconsistencies between clues and scraped data regarding ratings and download counts. It is suitable for the exploratory product stage when requirements are still only a single sentence. As a community-sourced skill, its quality varies with the version, and its effectiveness depends on whether the process is actually followed in response to the challenges.
Read MoreB2B AI Sales Assistant: Delegate Pre-visit Analysis and Preparation to Skill
The B2B Sales Assistant (v1.0.0, @user_fcba917f/tknintelligence) is a community-developed Skill under the business-ops category, designed to automate repetitive tasks such as meeting preparation. It provides six core capabilities: customer company analysis, industry trend insights, competitor tracking, customer pain point identification, meeting script generation with entry point recommendations, and personalized email drafting and sending. The customer company analysis function can access real business registry data by integrating with Qichacha and Tianyancha MCP services. This scenario-based Skill supports trigger keywords including sales analysis, customer analysis, meeting preparation, competitor analysis, customer pain points, meeting scripts, sales emails, send email to customer, customer insights, entry points, and configure enterprise data source. Tagged with requires_api_key=true, it requires API key configuration before use, with additional MCP service setup needed for enterprise business data. It is ideal for SDRs, Key Account Managers, and Sales Operations teams.
Read MoreTypographic Style Design for Print Publications | Jianshi AI: Delegating Publishing-Grade Typography Standards to Skill
Introducing the community Skill “Print Publication Typography Style Design | 简诗 AI” (design-media, @user_fccf428b/basic, v1.0.1). Targeting books, magazines, and reports, it offers a range of typographic variations from minimal to ornate and outputs visual specifications suitable for content publishing scenarios. It does not render files and requires no API Key. This article outlines its positioning, core features, installation entry and how to enable it (no usable installation command is officially provided; users need to confirm on the page), typical usage based on its positioning, as well as applicable scenarios and cautions.
Read More