Intelligent Essay Grading: Delegating Composition Correction and Writing Assessment to Skill
Essay intelligent grading and scoring is an education-focused large model skill launched by Kunlun Growth. It is located at @kunlungrowth/edu-essay-grader, version 1.2.1. It targets Chinese and English essays, and outputs scores, essay comments, and upgrade suggestions according to scoring criteria, covering grading of Chinese and English compositions in middle school and college entrance exams as well as daily teaching feedback. Trigger words include essay grading, Chinese composition, English essay, scoring criteria, writing evaluation, essay comments, and correction. The article explains its functional boundaries, triggering methods, precautions for activation, and the limitation of requiring verification of rubric suitability before grading formal exams.
Read MoreOne-Click Generation of Hand-drawn Comic Story Videos: From Story Text to 16:9 Finished Animation
This community Skill (@indiv-dfteacher/fmg-autoshouhui-vidio, v1.0.2, design-media) automatically converts Chinese story text or ordered images into finished 16:9 videos featuring a hand-drawn comic style. It first utilizes *story-to-handdrawn-video* to generate and render colored pencil diary-style storyboard illustrations, then invokes *fmg-autojy-video* to synthesize the final product with AI voiceovers, cover pages, background music (with auto-ducking to ensure voices remain clear), and subtitles. First-time usage guides the user through downloading local dependencies and deployment self-checks. The tag indicates requires_api_key=false. The article outlines what the tool is, its core features, how to enable it via coordinates, the official typical workflow, applicable scenarios, and precautions. Installation commands and specific parameters were not listed in the official documentation and have therefore not been fabricated.
Read Moredsh-soul: Independent Memory, Graph Compression, and Controlled Harness Evolution
DSH-soul 2.0.0 is a DSH workflow plugin maintained by icyaaaww. By default, it enables a memory skeleton (graph nodes, 600-byte projections, and soul_recall) and offers an optional graph compression backend to replace model summaries. Evolution is disabled by default; when explicitly enabled, it compiles experience into C1–C8 modifications, validated through pairwise verification, shadow trials, and checks acceptance. It requires Node.js 24+ and DSH ≥0.2.1-rc.2. Installation is performed using the `dsh plugin --profile` command pointing to a local tgz file. It is ideal for scenarios requiring long-session constraint retention, reduced compression costs, and controlled iterative harnesses.
Read MoreModeling Lottery Rules and Cross-Lottery Analysis: Upgrading Single-Lottery Experience to Cross-Lottery Consistency Verification
# Introduction Practitioners of lottery data analysis frequently encounter two categories of problems. The first is directly applying patterns from the "Double Color Ball" to the "Super Lotto," only to find that the patterns fail after switching lottery types. The second involves uncovering a multitude of "patterns" within a single lottery type, yet failing to distinguish between genuine structural patterns and small-sample noise. The common root cause of these issues is the lack of cross-lottery comparison: any pattern within a single lottery can be explained away as "lottery-specific noise." Only by testing these patterns for reproducibility across multiple lottery types can one determine their transferability. The Skill introduced below turns this process into a reusable analytical workflow. ## What is this "Lottery Rule Modeling and Cross-Lottery Analysis" is an Agent Skill designed for the analysis of lottery rules and probability structures. It is maintained by the community publisher user_04b42a16 and located at @user_.
Read MoreDrawing Review and Design Optimization Consultant – Brother Peng's Engineering AI: A Pocket Guide to Review Essentials and Value-Creation Cases
This Skill packages the accumulated drawing review checklists from regional subsidiaries of central state-owned enterprises (SOEs) and 237 optimization cases into a knowledge base of approximately 8,300 words. It covers review criteria and procedures, key points for nine professional disciplines, twelve categories of high-frequency optimization techniques, six major directions for value creation, and a risk boundary database. By simply inputting the project phase and discipline, users can generate drawing review logs, discipline-specific review checklists, optimization proposals, value-creation planning outlines, risk alert lists, and case study cards. It requires no data uploads or API keys. Please note: This tool does not replace design validation or official documentation. Optimization outputs involving mandatory codes, structural safety, fire protection, and specialized acceptance serve only as reference leads. Value-creation figures are derived from specific cases in the source data and must be recalculated for new projects.
Read Moredsh-media-desk: Complete Inspiration, Drafting, and Scheduling in the DSH Sidebar
dsh-media-desk is a community workflow plugin for DeepSeek Harness. It provides an inspiration wall in the sidebar, one-click generation of voiceover scripts, covers, illustrations, and storyboards, as well as content management, calendar scheduling, data review, and style presets. Data is stored in ~/.dsh/media-desk/. It requires DSH 0.2.x version 0.2.1-rc.5 or above. Download the tgz from Releases and install it using `dsh plugin add file:`. Text models use DSH configuration, while image generation and optional Monid account scraping require your own API keys. Open source under the MIT license, it cannot post content to platforms on your behalf.
Read Moredsh-marktext: Edit Markdown in DSH Right Sidebar Using MarkText Style
dsh-marktext, maintained by Roland0511, embeds @muyajs/core into the DeepSeek Harness sidebar, supporting WYSIWYG editing for *.md files, offline extensions like KaTeX/Mermaid, auto-save, and version guard write-back. Based on the official README, this article covers installation (in-app dsh-marktext, web profile CLI, and manual desktop package.json), the iframe architecture, diagnostic interfaces, and notes regarding mutual exclusivity with dsh-markdown-typora. Requires DSH 0.2.1+. MIT License.
Read MoreAI Large Model Expert | AI Image and Video Marketing for Renovation Companies: Introduction to AI Content Marketing Skills for Home Improvement Businesses
Introducing **@infimind/industry-renovation-company-marketing-ai-hive** (Version 1.0.0, Category: Design & Media) on SkillHub. Targeting homebuyers, tenants, homeowners, renovation clients, community residents, and project decision-makers, this tool centers on real-life cases, materials, craftsmanship, budgets, and construction progress. It generates audience and selling point strategies, 30-day content calendars, image asset matrices, short video columns and storyboards, itemized prompts, cross-platform adaptations, AI-HIVE task logs, and review metrics. Additionally, it facilitates AI-HIVE interactions by allowing queries for current image/video models, uploading reference materials with usage rights, saving price snapshots, routing based on cost/speed/success rate, polling for downloads, and logging taskIds. It is suitable for Douyin, Xiaohongshu, WeChat Channels, real estate and home furnishing platforms, WeChat Official Accounts, and private domain traffic. **Authenticity Boundaries:** Distinctions must be made between real-life footage and schematic illustrations regarding property listings, area, location, price, materials, construction duration, and effects; fabrication of projects is prohibited. No guarantees are made regarding viral success, sales volume, rankings, audit approval, or return on investment. **Tag:** `requires_api_key=true`
Read Moredsh-familiar-memory (dsh-neuron-memory): DSH Local Memory Layer
Elio896's dsh-familiar-memory (package name dsh-neuron-memory v0.2.0) provides local SQLite memory for DeepSeek Harness, featuring FTS5 literal recall, sentiment buckets with same-day context, exponential decay, and bottom-out archiving; vector and graph features are disabled by default. It achieves session continuity via neuron_opening/neuron_handoff, while the "work" scope uses supersedes to mark the latest decisions. Installation requires executing node install.mjs in the plugin directory and dsh plugin add link:<directory>, with configuration options for db, halfLifeDays, autoSync, etc. Data is not uploaded; automatic logging reads ~/.dsh/sessions/, so source code and permissions should be reviewed independently. MIT licensed third-party plugin.
Read MoreDesktop Pet: Replace the Pet Image with a GIF
"Desktop Pet" is a community Skill, located at @user_8fd82e77/zmcw-20260819, version 1.0.0, category dev-programming, publisher user_8fd82e77, with the brief description "Generate a desktop pet that can upload GIF images to replace related actions." It is positioned as a tool for changing a desktop pet’s appearance, using a v8 single-window composition architecture: it takes the user-submitted one idle-display GIF and several randomly played GIFs, removes the background onto a white base, generates project assets under E:\desktop-pet, registers the result as a "random playback every 5 seconds" effect, and repackages it into a desktop pet exe, so as to replace “布丁店长” with any anime/cartoon character design. It includes a self-contained background-removal script references/gen_swap.py, supporting two modes: light neutral-white and dark floodFill. A SkillHub page snapshot shows a rating of 100000, 1321 downloads, and 0 installations, with the tag requires_api_key=false. Installation follows the generic SKILL.md format by adding the Skill URL; there is no verified package-manager command. It is aimed at users who want to replace the appearance without manually modifying the project, and also works for collaborators to use independently.
Read MoreLet it be | Accuracy Check: A Structured Toolkit Skill for Traditional Numerology
"Forget It | Is It Accurate?" is a life-service Skill on SkillHub created by Cunlao Xie. It provides people who want to experiment with traditional divinatory arts with a computable, table-driven, framework-based toolkit composed of ten modules, covering Bazi (Four Pillars of Destiny) chart casting, Zi Wei Dou Shu, Qi Men Dun Jia, I Ching hexagram interpretation, Chinese zodiac fortune, Eight Mansions feng shui, name creation and name matching, tarot, personality-color analysis, and numerology. It addresses two specific problems: first, unified standards—built-in authoritative tables and algorithms lock in the casting rules and terminology for the ten systems to prevent inconsistent chart outputs from one generation to another; second, clear boundaries—it only presents structured results and does not output specific good/bad fortune predictions, and its results include a disclaimer that it is intended only for entertainment and traditional cultural research. The article introduces the positioning, a capability list for the ten modules, reproducible usage examples reconstructed from trigger words, as well as applicable scenarios, limitations, and pitfalls, including that it does not provide fortune predictions, that the built-in standards should be taken as authoritative, and that official documentation should be checked before integration.
Read MoreDesigning an Exercise Plan for Weight Loss: Transforming Fat Loss Goals into Actionable Exercise Prescriptions
"Creating an Exercise-Based Weight Loss Plan" is a community Skill on SkillHub, identified as @user_a549da14/jk-001 (Version 1.0.0, Category: life-service), published by user_a549da14. Centered on exercise prescriptions and supported by essential dietary calorie control principles, it addresses user needs such as weight loss through exercise, developing fat-loss training plans, calculating daily caloric deficits, selecting aerobic/strength combinations, and breaking through plateaus. It requires no API Key and is based on the general SKILL.md format. As the official page does not provide installation commands or configuration fields, the activation method depends on the Agent client's requirements for the Skill directory. It is suitable for handling targeted fat-loss inquiries within conversational products; however, the dietary section only covers principles of calorie control and does not replace professional nutritional or medical advice.
Read Moredsh-prompt-rewriter: One-click Rewrite of Drafts into Actionable Prompts Next to the Input Box
dsh-prompt-rewriter is a DeepSeek Harness community plugin maintained by huaguoxingkong. It provides a whale tail button next to the session input box that uses the current or a specified model to rewrite short drafts in-place into executable prompts containing objectives, formats, and boundaries. It only replaces the input box content without automatically sending it and supports Ctrl+Z and rollback. The introduction covers installation commands (npm / GitHub / Release), settings, context and image boundaries, shortcuts, and precautions, facilitating reduced multi-turn trial and error in model inference scenarios.
Read MoreAurabot (dsh-aurabot-qq): Integrating DSH AI as the QQ Bot "大肥鱼" for Chat
Aurabot (dsh-aurabot-qq) is a DeepSeek Harness client plugin that connects to the QQ alt account "大肥鱼" via SnowLuma or OneBot v11. It provides a full-page dashboard, with the master switch disabled and dry-run enabled by default, and offers modular capabilities covering private and group chats, personality/memory/mood systems, voice and emoji, Qzone, Bilibili, Douyin, Hexbox, and cover songs, with support for custom modules via a configurable data directory. Based on the official README, this article describes the architecture, installation commands, wizard steps, and Linux limitations, and provides typical usage examples and permission considerations.
Read MoreSelf-Improving + Proactive Agent (Pro Edition): Empowering Agents to Review, Self-Correct, and Accumulate Experience
Self-Improving + Proactive Agent (Pro Edition) is a community skill on SkillHub, located at @user_87f6522b/self-improving-proactive-agent, version 1.0.0, category ai-agent. It wires self-assessment, self-criticism, self-learning, and self-organizing memory into a fixed workflow that requires a one-sentence conclusion up front, delivery of a structured result file, and clear annotation of the evidence behind key decisions. Triggers include directly mentioning the skill name, asking how to do something, or asking the agent for help; there are five task scenarios: operation failure, user correction, outdated knowledge, discovery of a better approach, and explicit user installation or reference. The article outlines core capabilities, installation and activation methods, five reproducible typical use cases, as well as applicable scenarios and limitations. While writing, the official URL could not be opened; please refer to SkillHub for the installation command and page data.
Read MoreMoney-Finding Radar | Discover How You Can Make Money: Scan Your Personal Monetizable Assets into a List
**Money-Picking Radar** (@indiv-cunzhanglaoxie/asset-radar) is a community Skill categorized under business-ops, version 1.1.12, requires_api_key=false, created by Village Elder Lao Xie. It systematically scans your personal monetizable assets, using the four fundamental laws of logic (Law of Exchange / Law of Asymmetry / Law of Stock / Law of Minimum Closed Loop) to determine *why* something can be monetized, employs a five-dimensional scan (Skills / Inventory / Network / Time / Information Gap, five questions per dimension) to find out *what* you have, and ranks items using three valuation questions (Monetization Difficulty / Single-Transaction Value / Startup Speed) to output a Top 5 money-picking list, each item accompanied by this week's first step. Red lines: it only produces zero-cost startup plans—no income promises, refuses to scan gray-area information gaps, refuses to exploit acquaintance networks, refuses prerequisites of large upfront investment before monetization; if you want to scale up, switch to the OPC Business Model Designer—the Radar handles *finding*, OPC handles *scaling*. Enable via the official address and trigger with phrases such as "what can I do to make money," "asset inventory," "side hustle," "first dollar," etc.
Read Moredsh-startup-screen: Custom Boot Animation for DeepSeek Harness
dsh-startup-screen is a DSH plugin (v1.3.0, MIT license) by maintainer 6shenhonghong9 that plays a six-act startup sequence on cold boot: a black-screen logo, an overexposed white-out, a warm-white archive backdrop, a five-step access authorization, and a chromatic-dispersion reveal, with support for interactive confirmation, skip, real audio/TTS, a liquid-glass style, and warm-white/dark themes. Install via DSH Settings → Plugin Marketplace, or by running: dsh plugin --profile web|desktop add <local-tgz>. After installation, restart and tune 30+ parameters under Settings → Startup Animation. Intended for fun theming and personalized openings; review the source code and third-party asset disclaimers before installing.
Read MoreEight Characters (BaZi) Chart Calculation Skill: Deriving Four Pillars of Destiny (Stems and Branches) from Birth Time
The Bazi Charting Skill (@user_1854b633/bazi-paipan, v2.1.0, community source, life-service category) calculates the Four Pillars (Bazi) based on birth time, location, and gender. It outputs hidden stems, Ten Gods, the Twelve Stages of Existence, Void (Kong Wang), Na Yin (Musical Tones), Five Elements statistics,刑冲合害 (punishment, clash, combination, harm), and auspicious/inauspicious spirits (Shen Sha). It generates 10 steps of Major Luck (Da Yun) and supports specific years for Annual Cycles (Liu Nian). It features True Solar Time correction and the VSOP87 solar term algorithm, requiring no third-party dependencies. As indicated on the page (requires_api_key=false), no API Key is required. The official description explicitly states that it only explains what is on the chart, how to interpret the terminology, and the source books. It does not make judgmentments on good or bad fortune, provide conclusions on pattern quality or favorable/unfavorable gods (Xi Yong Shen), offer advice on luck-changing remedies, or collect personal information. Trigger words cover Bazi, Four Pillars, Charting, Ten Gods, Hidden Stems, Na Yin, Major Luck, Annual Cycles, Zi Ping, and Destiny/Mingli. The installation command was missing from the retrieved data and remains unverified; the text states this honestly. It is suitable for developers requiring stable charting output or terminology explanation documentation. Please note that different schools have different calculation methods; this is for reference only and does not constitute a basis for decision-making.
Read MoreHAP Skills: A Collection of Skills for Direct AI Manipulation of Mingdao Cloud HAP
hap-skills is a collection of Mingdao Cloud HAP AI Agent skills maintained by community publisher user_175d84c3. Version 1.0.1 includes 9 skills organized into three categories: CLI (hap command line), MCP (HAP MCP service), and API (HAP V3 HTTP interface). It adopts a master entry plus sub-skills structure: the root SKILL.md serves as a unified entry point, automatically routing by intent to sub-skills for creating apps, modifying apps, querying data, building websites, creating interfaces, and developing plugins. It supports both full-suite installation and on-demand individual installation. hap-cli handles installation, login, and navigation for contacts, chat, schedules, and approvals; hap-cli-app-creator enables one-stop app creation; hap-cli-app-editor facilitates partial modifications to fields, views, permissions, and workflows; hap-cli-data-query manages multi-condition filtering and pivot aggregation; hap-cli-environments manages multiple environments and accounts with confirmation before destructive operations; hap-mcp-app-builder allows conversational full-automatic app creation with support for interruption and resumption; hap-apiv3-data clarifies authentication and CRUD operations; hap-api-website guides the construction of decoupled front-end and back-end websites; hap-view-plugin covers the entire development workflow for custom view plugins (mdye). SkillHub rating: 100,000; Downloads: 1,564; Tag: requires_api_key=false.
Read MoreOmnichannel Customer Acquisition Skill: Transforming Customer Search into an AI-Outsourced Funnel Pipeline
Targeting physical store owners, knowledge-pay and consulting professionals, and enterprise sales teams seeking stable customer leads, this article introduces the "Omnichannel Customer Acquisition | Effortless Traffic Generation" Skill on SkillHub. It defines customer acquisition as a funnel pipeline: Reach → Hook → Trust → Lead Capture → Handling → Closing. The workflow utilizes AI for batch material and script generation, while humans focus on final decision-making and closing deals. The Skill covers audience insights, a hook library, topic and script templates, short video and livestream SOPs, Xiaohongshu (Little Red Book) SOPs, private domain lead handling and conversion, search and local area acquisition, case study breakdowns, data reviews, prompt packs, paid advertising SOPs, cross-border B2B/B2C acquisition, and the generation of posters, PPTs, storyboards, and cover materials. Featuring a built-in one-click generation engine, it produces a "three-piece set" upon entering industry keywords or simple commands like "I need to generate leads" or "Write a short video script." It includes a library of 20 high-frequency scenarios and automatically identifies ToB/ToC distinctions, persona types (persona vs. informational content), and business structures (corporate vs. solopreneur). The article also outlines the target audience, trade-offs in multi-channel synergy, and limitations of automatic judgment. It transparently notes that the official page content could not be scraped and the installation commands could not be verified.
Read Moredsh-unrestricted: Injecting switchable restriction-bypass rules at the DSH system prompt stage
dsh-unrestricted is a DeepSeek Harness plugin maintained by yexi-by. It injects an unrestricted:execution-mode rule block into the system prompt following the official role settings within the system-prompt/assemble pipeline, and provides a real-time toggle via cordis.patch.yml within the profile. It supports Standard, Cordis, PTC, Plan, Minimal, and sub-agent configurations, with the settings page allowing preview of the complete system prompt for each mode. Installation via the official command for v0.2.7 requires enabling both the plugin and the "Enable Unrestricted Mode" option. It does not bypass API gateway filtering or the DSH sandbox. Open-sourced under the MIT license.
Read MoreContent Strategy：把内容当作一套可复用的生产系统
Content Strategy（@clawhub_jk-0001/content-strategy，v0.1.0，ClawHub 内容营销类 Skill）面向个人创业者，把内容生产的前置环节整理成八步：定目标、挖选题、定支柱、选格式与渠道、排日历、做分发、内容复用、看指标。文章按 SKILL.md 官方内容逐节说明，给出 openclaw skills install 安装命令、n8n 顾问的支柱与日历示例、一条长内容拆成五类短内容的复用流程，并说明适合场景与局限；同时指出第三方页面上的主题集群、60/30/10 配比、四维选题打分等描述与官方资料不符，未予采用。
Read MoreNuwa.skill (nuwa-skill): Distilling Anyone's Way of Thinking in DSH
Nuwa.skill (alchaincyf/nuwa-skill), based on the Agent Skills protocol, automatically researches and distills expression DNA, mental models, decision heuristics, and boundaries of honesty by simply inputting a name or topic, thereby generating a runnable Skill. It can be installed in DeepSeek Harness via `dsh plugin add github:alchaincyf/nuwa-skill`, with the distillation results written to the project's `.dsh/skills/` directory. This article introduces the five-layer distillation structure, official Skills for existing personas and themes, installation and uninstallation commands, and typical conversational usage, while advising users to review the source code and license before running it with the current DSH process permissions.
Read Moreai-video-script: Turning Topics/Keywords into Storyboard-level Video Scripts
This article introduces the Skill ai-video-script (@aguo333/ai-video-script, v1.0.0, published on ClawHub, categorized under content-creation, no API Key required). By inputting a theme or keywords, it outputs a complete script including a video overview, storyboard, AI visual prompts, voiceover script, and subtitles. The article explains the generation workflow, prompt specifications for both CogVideoX and Midjourney, advanced features (such as batch generation, style customization, and duration optimization), the OpenClaw installation command, as well as applicable scenarios and considerations regarding copyright and platform compliance.
Read MoreInteractive Courseware | Qiuye Workplace: Developing a Single-File Interactive Courseware for Offline Use
Interactive Courseware_Qiuye Workplace Community Skill, located at @user_c774f8e9/interactive-courseware, version 1.0.2, categorized as education, tagged with requires_api_key=false. It organizes the teaching content of a lesson into a ready-to-use interactive courseware: it first defines teaching objectives and key difficulties, then breaks down knowledge points into interactive tasks such as drag-and-drop matching, multiple choice, ordering exploration, scenario simulation, and instant feedback. It features navigation, progress tracking, timing, and real-time comments. Finally, it checks against three criteria for courseware evaluation and outputs an HTML interactive courseware (single file), a Word teaching design and usage manual, and an Excel list of interaction points and assets. It applies to scenarios such as interactive courseware, H5 courseware, classroom interaction design, open class and competition courseware, in-class quizzes, in-class exercises, and as an alternative to Seewo courseware. The retrieved SkillHub page shows a rating of 100000, 1747 downloads, and 0 installations. As the current materials do not provide specific installation commands or the content of SKILL.md, the article states this factually and suggests referring to the official page for details.
Read More