Online Shopping Price Comparison Tool: Real-time Cross-platform Comparison of Similar Items
# An Agent Skill Based on SKILL.md: Online Shopping Price Comparison Tool **(@user_899dd41d/shopping-price-comparison-tool, v1.4.0, Community Release)** This article introduces an Agent Skill designed for cross-platform price comparison scenarios. The tool covers Taobao/Tmall, JD.com, Vipshop, and Douyin Mall, aggregating data including prices, sales volume, historical lowest prices, national subsidy policies, coupons, free shipping information, user reviews, as well as authentic evaluations from Xiaohongshu (Little Red Book) and Douyin. It generates visualized HTML comparison reports with bar charts. Key features include product model fingerprint recognition to prevent confusion between similar models, refined filtering for Taobao/Tmall and JD.com self-operated stores, JD logistics labeling, as well as scheduled price monitoring and push notifications. The article outlines its positioning, verified core functions, installation and activation methods (based on the official SKILL.md), applicable scenarios and precautions, and provides the official address.
Read MoreData Visualization / Scientific Data Visualization: Automatic Binning and Chart Recommendation for Histogram Data
This community Skill (`@user_f4fc8899/bio-data-visualization-integrated`, version 1.5.7, category data-analysis) streamlines histogram data preparation into a single call. By accepting JSON numerical data as input, it automatically performs binning, calculates frequencies and cumulative distributions, and recommends chart types. It runs via a hosted executor, returns structured content (providing Markdown when configuration is involved), and neither runs upstream local toolchains nor manages third-party accounts. It performs data preparation only and does not generate graphics. It requires no API key, and the billing method is Alipay. Installation instructions are lacking in the provided materials; therefore, users should verify the official installation guidelines before use, as the official page failed to retrieve during this session. Conflicting popularity data was observed: one retrieval showed a rating of 100,000 with 855 downloads, while another showed 99,747 and 787, respectively; please refer to the live page for accurate figures.
Read Moredsh-mcp-auto-permission：工作区写入 + MCP 预设与 MCP 审批自动认领
dsh-mcp-auto-permission is a DSH bundle maintained by ThomasTZh. It registers the workspace-write-mcp preset and, under that mode, claims mcp__* approvals as allowed-once, without modifying the sandbox or hard deny rules. This article describes its behavior, approval/request hooks, installation link steps, the autoApproves interface, and limitations such as headless mode and sub-agents. Review the source code before installing.
Read MoreCryptography Architecture Design: A Skill on Demystifying Key Management, End-to-End Encryption, and TLS Configuration
Introducing the "Cryptographic Architecture Design" Skill on SkillHub, located at `@org-02qudk26/crypt-zh`. Published by Zhiyu Technology from the enterprise source, it is categorized under `it-ops-security` and is now in version 1.0.0. During the design phase, it covers six core capabilities: algorithm selection, key management, end-to-end (E2E) encryption, KMS integration, signature verification, and TLS configuration. It is specifically tailored for the design of cryptographic protocols, key rotation workflows, and E2E encryption architectures. This article explains its positioning, core capabilities, enabling methods, typical usage scenarios, and applicable boundaries, and lists its official URL.
Read Moredsh-chat-preset: Adding Restricted Chat Presets for DeepSeek Harness
dsh-chat-preset is ThomasTZh's DeepSeek Harness bundle, registering an agent preset with the id `chat`, which provides web_search, web_fetch, and skills. Through a guard, it restricts tools after the agent is created to prevent reading/writing the workspace and executing shell commands. The article covers the preset's fields, the design rationale for the guard, the installation method via official clone and profile dependencies, as well as verification items and limitations.
Read Moredsh-web-search-brave: Integrating Brave Web Search into the DSH Official ctx.web Seam
The `dsh-web-search-brave` plugin, maintained by `thomasly`, registers the Brave Search API into the `ctx.web` seam of DeepSeek Harness 0.2.x, with the provider ID `brave-search`. The plugin has zero dependencies, does not import host packages, and reuses the `BRAVE_API_KEY` from `~/.dsh/.credentials.yaml`, enabling models to connect directly to Brave via `web_search`. Based on the README, this article explains the installation command, how to enable it via `cordis.patch.yml`, optional configuration, and the key resolution order. It also reminds users that the bundle must be restarted for mounting, and that the plugin accesses the network with the permissions of the `dsh` process.
Read Moredsh-skill-easy: Categorize DSH Skills and Localize `/` Menu Descriptions to Chinese
dsh-skill-easy is a DSH workflow plugin maintained by ther000. It requires DSH 0.2.1-rc.3 or later. Users can create their own categories under “Settings → Skill Categories” and drag in the skills. Each skill can have its Chinese description entered manually, or the host’s translator can be invoked with one click to batch translate them. The `/` menu displays the Chinese descriptions grouped by category. Data is stored in ~/.dsh/storages/dsh-skill-easy/data.json. It can be installed from the marketplace or from a local directory by instructing the assistant in chat (e.g., “Install the dsh-skill-easy plugin”); after installation, a restart is required. If host UI changes affect menu enhancements, the plugin can automatically degrade gracefully while still retaining the user-created categories and Chinese data.
Read MoreRuizhiheng Distributed PV IRR Calculation: Let the Model Handle Distributed PV Returns Estimation
This article introduces the community skill on SkillHub: "Ruizhiheng Distributed PV IRR Calculator for You" (@user_40df3c9e/pv-irr-calculator, version 1.0.3, categorized under business-ops). It streamlines distributed photovoltaic (PV) revenue calculation into a single conversation: simply input the city, installed capacity, and electricity price to receive the total investment IRR, NPV, and payback period directly. Built-in solar resource data for 33 cities and coal-fired benchmark prices are included, supporting dual revenue models (self-consumption plus surplus power feed-in), comparison of five consumption rate scenarios, and 25-year year-by-year cash flow analysis (with module degradation and electricity price escalation included). The skill is suitable for investment estimation of commercial and industrial rooftop PV systems, economic analysis in project proposals, and campus solution comparisons. It is tagged with requires_api_key=false, meaning no API Key is needed, and can be triggered by keywords such as "PV IRR" or "commercial and industrial PV calculation". The official page shows a rating of 100,000, with 879 downloads and 0 installations.
Read MoreCodeFlow Canvas: Turn Agent-Driven Coding Tasks into Auditable, Rollback-Capable Workflow Diagrams
CodeFlow Canvas (dsh-codeflow-canvas 1.4.2) is a DSH desktop workflow plugin maintained by TheLiudawei: nodes represent coding tasks executable by agents, and edges represent data dependencies. It supports /flow-import, a fully deterministic local scan that builds both the task graph and the module graph without generating any LLM tokens, a dual-view layout, execution tracking, and per-run undo of the state restored from the snapshot taken before RUN. Based on the official README, this article introduces the tool's positioning, its differences from transcript-based planning, the installation command, the typical entry points, and the sessionController prerequisites, while reminding readers to review the source code and requested permissions before installing.
Read MoreAI Songwriting Engine: decomposes songwriting into a reproducible pipeline
The AI Song Creation Engine is a design-media Agent Skill developed by Douxin Information Technology Studio, located at @douxin-tech/ai-song-writer. It breaks down songwriting and composition into a 9-step pipeline: Core Sentence First → Emotional Landscape Map → Melody DNA → Groove Profile → Harmonic Narrative → Lyric-Melody Interlocking Check → Tension Audit → Rubin Subtraction → Six-Dimensional Review. It also covers four Genre-specific sub-pipelines: Pop, Rap, Folk, and Electronic. The process is based on the creative principles of 50 songwriters, including McCartney, Leonard Cohen, and Rick Rubin. Starting with Version 2.4, it supports one-click audio demo generation while outputting dual files in .mid and .html formats. The HTML file features a built-in Web Audio player, allowing users to hear chord progressions, melody lines, drum kits, and lyrics synchronized simply by opening it in a browser; the MIDI file can be directly imported into a DAW for continued arrangement. The output of a complete run includes a Creative Core Card, Emotional Landscape Lyrics, Sound/Arrangement/Groove/Harmony details, Review and Rubin Subtraction conclusions, a verdict, JSON metadata, and an audio demo. It is suitable for scenarios such as songwriting, lyric writing, hook design, melodic motif development, arrangement reference, short video scoring, and digital human songs. The output is an AI-generated preliminary draft of lyrics and music, free for personal use. However, commercial distribution requires full music copyright registration/filling. Replicating the voice of real persons or plagiarizing mature works is prohibited. The documentation states `requires_api_key=false`, meaning no additional API Key is required. The text faithfully notes discrepancies in version records: the fetched data is labeled as Version 1.0.3, the introduction describes features of Version 2.4, while the page display shows the latest update as v1.0.2 from 3 weeks ago.
Read MoreHarvard's Science and Engineering Learning Method: Deconstructing "Rapid Industry Comprehension" into a Three-Step Cognitive Framework
The Harvard Tech Learning Method (@user_02d8b6c9/harvard-tech-learning) is a community-sourced Skill categorized under education, versioned at 1.0.0, and tagged with requires_api_key=false. To address the issue where open-ended questions like "Help me understand the XX industry" yield only vague, generalized overviews, it structures the process of quickly learning an industry, domain, or discipline into a three-dimensional cognitive framework: industry consensus, controversial issues, and in-depth verification questions. Triggered via natural language commands, the officially listed trigger scenarios include "Help me understand the XX industry," "introduction to the XX field," "grasp XX quickly," and "Analyze XX using the Harvard Tech Learning Method." The article explains its role and three core capabilities, provides ready-to-use sample prompts, and highlights its limitations: it only provides the structural framework of questions rather than drawing the conclusions themselves; if contextual information about a domain is insufficient, the framework may struggle to adequately fill out all three dimensions. Furthermore, the metrics—100,000 ratings, 765 downloads, and 0 installs—were obtained via web scraping from the official page records. At the time of writing, both the official page and the API address failed to respond, and the installation command could not be verified; thus, speculative installation commands are omitted.
Read MoreWang Yangming Philosophy-Driven Action Coach for Work, Management, and Life: Turning Stuck Issues into Actionable Next Steps for Retrospection
Anchored in Wang Yangming’s principles of the unity of knowledge and action, practice through real affairs, resolve, introspection, correcting faults, and offering responsibility to help others, this Skill integrates modern practices—such as Naval’s insight, Drucker’s management, lean methods, habit design, decision science, post-action review, and organizational learning—to help everyday people navigate career and life challenges. It is not a role-play persona or a vault of quotable lines; it is an action-coaching workflow. Users describe a recent issue where they feel stuck. The Skill first identifies the true problem, then creates immediately actionable micro-steps, observation metrics, and stop conditions. Users can return later for a review cycle, and with their confirmation, plans are saved as local Markdown for the next session. The thirty built-in prompts serve only as content anchors; for new issues, the Skill performs bounded reasoning guided by the School of Mind’s logic. It makes no promise of promotions or financial gain, provides no investment, medical, or legal advice, does not assign a “conscience score,” never stores records without explicit user approval, and never forces issues into a predetermined interpretation.
Read MoreB2B Sales Strategy: Standardizing the Production Process of Strategic Reports for Government and Enterprise Key Accounts
**Advisory sales for government and enterprise key accounts** present challenges due to inconsistent standards in generating combat sales strategy reports. Reports produced by different consultants vary significantly in customer analysis depth, entry points, and the pace of progression. The B2B Sales Strategy Skill (@user_9a8549a5/b2b-sales-strategy, version 1.0.0, community-sourced under the "business-ops" category) generates enterprise-level B2B sales combat strategy reports based on a six-dimensional integrated model. It supports direct input of customer information or conducting interactive research guided by the SPIN and strategic sales frameworks. It applies to digital projects and key government and enterprise accounts involving AI, data, software, software-as-a-service (SaaS), and requires no API Key. As the developer has not provided detailed installation commands, it must be loaded via coordinates in tools supporting the SKILL.md format. Given its community-sourced nature and zero installation count, it is recommended to conduct a small-scale pilot validation prior to official adoption.
Read MoreConnect to WeChat (dsh-wechat-bridge): Enabling Two-Way Chat between DSH AI Partners and WeChat
dsh-wechat-bridge is a DSH client plugin maintained by thawline (based on the upstream project by Yuuyuko-uu). It integrates WeChat conversations into DSH sessions by reading messages via read-only access to the local WeChat 4.x database and sending messages via Windows UI Automation. It supports group chat rules, owner privileges, stickers, and group diaries. Installation: dsh plugin --profile desktop add github:thawline/dsh-wechat-bridge. Windows only. There are risks associated with automating a personal account. Please review the source code and the MIT license before use.
Read Moredsh-phone: Running Backend Android Instances in the DSH Right-hand Sidebar
dsh-phone is a client plugin for DeepSeek Harness (v0.4.1, MIT), which runs Android in the backend (redroid / SDK Emulator / attach), displays the scrcpy screen in the right sidebar of the browser, and registers nine tools—such as phone_status, phone_screenshot, and phone_tap—with the Agent. To install, download the tgz from the GitHub Release and run `dsh plugin --profile web add`; do not use the similarly named npm package. It supports lazy startup, lifecycle management, and one-command attach integration. Requires DSH >=0.2.1-rc.2 <0.3.0 and Node >=20.
Read MoreGovernment Procurement Bidding Decision: Evaluation of Government Bids — A Decision Analysis Assistant for Go/No-Go Bidding.
Government procurement bidding decision analysis assistant, maintained by community publisher user_a15e4142, version 1.0.9, categorized under business-ops. When provided with a specific government procurement tender project, it outputs recommendations on whether to bid, identifies restrictive signals (invitation bidding / single source / tailored clauses / extremely short bidding windows), analyzes the procurement unit's supplier landscape and preferences, forecasts competitors, provides historical transaction prices for similar projects and suggested quotes, and assesses disqualification red lines and compliance risks. It generates decision reports based on historical procurement and bidding data from across the web. An API Key is required for configuration. No installation command was provided in the source materials; please refer to the official page for details.
Read Moregeo-case-generator: Rapidly Generate Practical GEO Optimization Case Reports by Industry
geo-case-generator is a community-sourced content-creation Skill (@user_1831d4fb/geo-case-generator, version 1.0.0, no API Key required). It rapidly generates practical GEO (Generative Engine Optimization) case study reports based on an enterprise's industry, scale, and pain points. It covers over 20 industries, including mechanical equipment, education and training, building materials and home decoration, business services, chemical energy, packaging and printing, and hardware tools. It is applicable to scenarios such as generating GEO case studies, industry case analysis, and customer case demonstrations. The article introduces its positioning, core features, installation and activation methods (specific commands are not provided in official materials, the page is currently inaccessible, so it is based on pre-fetched data), typical input/output usage, and usage limitations: report quality depends on the specificity of the input, effectiveness data and customer names need to be verified by the user, and the generated content is a draft rather than a real case study.
Read Moredsh-emotion-analyzer: Local ONNX Emotion Scoring and Session Hover Monitoring
dsh-emotion-analyzer is a DSH model-inference plugin maintained by thatgfsj-dsh (v0.2.2, requires DSH >= 0.2.1-rc.2). It performs local ONNX INT8 scoring on each assistant reply from the main model, projects the results into the Session’s emotion dimension via mapping, and injects a draggable floating box on the Web UI. It defaults to the horizon-small model (36 languages, 28 labels) and falls back to legacy behavior when no model is present. Inference runs entirely offline and does not call any third-party APIs. To install, run `dsh plugin --profile web add <local-path>` and verify with `check-install.sh`. Models can be downloaded and selected via the settings page or the `/api/emotion-analyzer/v1/*` endpoints. It is suitable for locally observing conversational emotion trends within the Harness. For commercial use, be aware that the legacy model is licensed under CC-BY-NC-4.0.
Read MoreData Insight, Visual Analytics, Intelligent Conversion: Turning Data, Files, or URLs into Structured Insights
This article introduces a community-released Skill: “Data Insights, Visualization Analysis, and Intelligent Conversion” (coordinates @user_2fd890c9/pretzelai, version 1.0.2, category: data-analysis). The core capability of this Skill is to transform data, files, or URLs into structured insights and visualization results. The tag requires_api_key=false indicates that no additional API Key configuration is required. The article lists its basic attributes, including source: Community, publisher: user_2fd890c9, rating: 100000, downloads: 828, installations: 0, and notes that the install count of 0 implies there are no public usage records. As the official page could not be accessed successfully this time, installation commands, directory paths, configuration fields, and reproducible examples are missing from the available material; the article does not fabricate these details and recommends verifying them on the official page before proceeding. It also reminds readers that the source is the community and the version is relatively new, so compatibility should be confirmed before inputting complex files.
Read MoreSmart Tutoring (Error Log): A Skill for Parallel Processing of Photo-Based Explanations, Error Log PDFs, and Bilibili Videos
Smart Tutoring (Error Journal) is a community-published skill coordinated at @user_36ddf2ef/xsd-smart-teacher and maintained by user_36ddf2ef. Currently at version 1.0.3, it is categorized under education with a community source. Tagging indicates `requires_api_key=false`, meaning no pre-configuration of API keys is necessary. Upon the user uploading photos of homework or exam papers, a multimodal model automatically identifies the questions and provides step-by-step explanations. It synchronously generates interactive, visualized HTML explanations (primarily targeting Math and Physics). Meanwhile, it asynchronously calls MCP to generate a PDF error journal and searches Bilibili for relevant explanation videos based on the questions' knowledge points. All three products are generated simultaneously and presented together in a parallel pipeline. MCP authentication dynamically checks for the existence of the xsd-mcp connector; if absent, it guides the user to add it manually (without proactively writing configurations). It prioritizes OAuth (WorkBuddy automatic handshake, 30-day authentication-free access) and falls back to a device code flow to issue a 30-day JWT. Trigger keywords include "error question," "error journal," "homework," "exam paper," "explain question," "explanation," "this question," "photo explanation," and "grading." This article introduces the Skill's core features, installation and activation methods, typical usage flows, and applicable scenarios.
Read Moredsh-auto-vision: Enabling Pure Text Models to Automatically "See" Images Under the Same Provider
dsh-auto-vision (npm: dsh-auto-sight) is a DSH model inference plugin maintained by thanderming. It intercepts requests containing image blocks on llm/stream where the main route does not support vision. It employs a vision model on the same Provider to read the images, replaces them with text descriptions, and then passes them to the selected text model for a response. It does not store API Keys and allows specifying the vision model and timeout via optional Settings. Installation: `dsh plugin add github:thanderming/dsh-auto-vision`. It is suitable for users who have already configured a single Provider and do not wish to set up a separate vision API; however, each uncached image incurs an additional call, and it is strictly limited to the same Provider.
Read Moredsh-voice-unified: DSH Unified Voice Plugin (Local Multi-Voice, Emotional TTS & Cover)
Ten-feng's dsh-voice-unified is a DSH model inference plugin (v1.6.2) that centrally manages multi-voice profiles, incremental training, seven-emotion TTS, cover song voice conversion with direct output, and music cache integration locally, with no audio uploaded by default. It supports a Ctrl+Shift+V management panel, on-demand/always-on voice replies, and WeChat-style bubble playback; DeepSeek side invokes 14 commands via the voice entry and underscore tool aliases. Installation requires Windows 10/11 x64 and DSH≥0.2.1; after adding the plugin using the link method, restart DSH.
Read MoreAI Image Analysis and Understanding | Jianshi AI: An MCP Skill for Image Understanding Using MiniMax Multimodal Models
This Skill is published by user_fccf428b from the community. Version 1.0.1, categorized under ai-agent, it is based on the MiniMax multimodal model and utilizes MCP tools for image description, recognition, and content understanding. The tag requires_api_key=true. The article outlines its positioning, verified core capabilities, activation conditions, and applicable boundaries, noting that official documentation does not provide installation commands or reproducible examples.
Read MoreOne-Click Short Video Generation: Turning Text and Media into Final Videos
This article introduces SkillHub's "Short-Video Script to Finished Video" One-Click Skill, located at @infimind/short-video-script-to-finished-video. Published by JRK Tech - AI Expert and currently at version 1.0.0, this enterprise-sourced tool falls under the "content-creation" category and requires an API Key. Core capabilities include generating videos from scratch using text, multimodal control over images/videos/audio, frame-to-frame animation, video editing and extension, as well as automatically uploading assets, saving tasks, querying progress, and downloading finished videos via AI Hive. It covers scenarios such as AI advertising, TVCs, e-commerce videos, product showcases, livestream commerce, viral product recommendations, micro-dramas, motion comics, and social media content. It also serves as a viable alternative to tools like Kling, Dreamina, Hailuo, MOKI, Vidu, PixVerse, Runway, Pika, Sora, Veo, CapCut, and HeyGen. The article outlines its installation and activation methods, typical input approaches, and key considerations, including reliance on external generation services, API Key configuration requirements, and the fact that it currently has zero installs to date.
Read MoreQClaw-CN Replacement Migration: AI-HIVE Multi-Model Workflow—Run a Small Sample First, Then Decide the Migration Scope
This is a migration-assessment Skill for developers and business teams, located at @infimind/ai-hive-product-switch-qclaw-cn, version 1.0.0, published by Jirui Technology-AI Expert, sourced from enterprise, categorized as dev-programming, and tagged requires_api_key=true. It triggers when users search for QClaw-CN alternatives, replacements, migration, similar tools, or API alternatives. It works by breaking the current real-world task into comparable samples, querying the AI-HIVE MCP service for the models, fields, pricing, and limits available that day, first completing a small code-generation migration sample, and then deciding which generative-AI workflows to migrate and which features to retain. The provided materials explicitly state that AI-HIVE does not claim any partnership with QClaw-CN, and do not reproduce its proprietary interface, data, community, hardware, or account assets. The captured materials do not provide installation commands, directory paths, MCP integration methods, or configuration fields, so those items are left blank here and marked as subject to the official page. The official page lists a rating of 100000, downloads of 1091, and installs of 0; since the highlighted data is inconsistent with this, the official page is treated as authoritative.
Read More