Generative Media Skills for AI Agents: Enabling Agents with Generative Media Capabilities
This article introduces the skill "🎭 Generative Media Skills for AI Agents" (ID: @org-eyw2ohcx/samuraigpt-generative-media-skills, version 1.0.0, enterprise source, publisher: RealQ, category: design-media) available on SkillHub. The skill wraps generative media calls in a standard SKILL.md file, powered underneath by MuAPI, and is tagged with requires_api_key=true, meaning an API Key must be configured before activation. Because publicly crawled materials did not include the full SKILL.md body or installation commands, this article only states verifiable facts and explicitly notes that the interface list, installation methods, and call examples must be taken from the official page; no speculation is made. The official URL and the rating and download data captured at the time of crawling are provided.
Read Moredelivery-video: Turning working features into verifiable delivery demos
The DSH workflow plugin `delivery-video` for Lumos-789 transforms functional prototypes into SHA256-bound delivery demos. It features verify-before-record logic, `episode.json` recipes, edge-tts narration with word-synced subtitles, and strict preflight/render/verify plus render gate enforcement. Install via `dsh plugin add` with a local path; requires python3, ffmpeg, and edge-tts>=7.3.1. It also supports feedback for defect videos. MIT 1.1.1.
Read MoreHong Kong CBBC Outstanding Risk: Query on Distance to Trigger Price and Outstanding Volume
The Skill from Antseer MCP, based on Futu OpenAPI v10.7, provides the ant_hk_stock_data tool, covering three categories of Hong Kong stock data: the call price and market concentration of callable bull and bear contracts, the spot prices of the underlying stock and the Hang Seng Index, and K-line backgrounds, with native Hong Kong dollar (HKD) quotations. It is suitable for distance monitoring and market concentration analysis of callable bull and bear contracts. Installation commands and configuration fields are missing from the existing documentation, and no official page was verified in this review; the tool must be enabled strictly based on official URLs.
Read MoreShort Video Editing Pacing and Final Video Quality Checks ｜ Jianshi AI: Transforming Final Video Quality Checks into an Actionable Checklist
An Agent Skill designed for short video and livestreaming teams, it handles editing rhythm control and final video QA across topic selection, production, publishing, and post-mortem review for Douyin, WeChat Channels, and Xiaohongshu. It delivers actionable plans, checkpoints, and risk boundaries. The article outlines its positioning, core features, installation and activation methods, applicable scenarios, and precautions.
Read MoreBaton: Deterministic Scheduling of Tickets and Subagents for OpenSpec Changes
Baton (dsh-openspec-baton) is a DSH workflow plugin for Lumos-789: it deterministically schedules tickets/DAGs on approved OpenSpec changes using SessionController and the session_control.py CLI. The main session forwards actions to subagents, verifying success strictly via sealed receipts. It supports per-task isolated git worktrees, up to 5-way parallelism, and a live tasks.md plan. Requires DSH 0.2.1-rc.2+, Python 3.10+, and a subagent tool; install via `dsh plugin add <local_path>` or `dsh-openspec-baton`.
Read MoreImage to PPT: Batch convert image folders into one-image-per-slide PPTX files
This article introduces the community Skill "Images to PPT" (@user_6fd83c94/images-to-ppt, Version 1.0.0, category: office-efficiency), which converts image folders into PowerPoint (.pptx) presentations with a "one image per page" layout. Each image occupies one slide, fills the slide completely while maintaining aspect ratio and centered alignment, supports png/jpg/jpeg/bmp/gif/webp/tif/tiff.Media, and requires no API key. It explains titles positioning, core functionality, and trigger use cases (e.g., upload an image folder to generate a deck, batch-convert images to slides, or one-image-one-slide workflows), and clarifies that it handles image placement on slides and does not manage text layout slide ordering depends entirely on filenames. The installation command is not included in the available documentation and has not been fabricated—readers are directed to the official address for installation instructions
Read MoreTransforming Intensive Paper Reading into Structured Reports
Literature Close Read is an Agent Skill tailored for academic paper reading scenarios, located at @user_9cc4d736/literature-close-read. It is version 1.0.0, maintained by the community publisher user_9cc4d736, categorized as professional, and does not require an API key to enable. It does not retrieve literature or download full text; instead, it performs structured interpretation of provided paper content and outputs a report covering the research topic, research design, methods, results, innovations, limitations, gaps, and overall research logic. It supports three usage modes: in-depth reading of a single paper, comparative reading of multiple papers, and extraction of methods and/or results. The article outlines its functional boundaries, how to enable it, reproducible usage examples, and cautions: report quality depends on the completeness of the input text; when method details and data are involved, the source text should be consulted for verification.
Read Moredsh-session-tree: Visualize Workspace DSH Sessions as a Browsable Fork Map
Session Tree (`@lulaalu/dsh-session-tree`) is a DSH 0.2.x client plugin maintained by lulaalu. It opens a full-page "session tree" via a tree entry in the web sidebar. Acting as a read-only interface for `ctx.sessions` and `turnOutline`, it visualizes sessions, conversation turns, and forks as a map organized by workspace. It supports history inheritance, sub-sessions, archiving, and node navigation. This article covers the design philosophy, features, installation methods (via GitHub, npm, and indexed sources), and known limitations. Please verify the source code and MIT license before installation.
Read MoreSuper Lotto Statistical Prediction System v1.1: Breaking Number Generation into Verifiable Statistical Processes
**# Introduction** If you tell a model "help me predict Super Lotto numbers," in most cases, it will either make up a set arbitrarily or slightly modify the numbers from the previous draw. While the numbers are generated, you cannot determine the basis for them, nor can you verify afterwards where the error occurred. Professionals in data analysis are accustomed to a different approach: every step relies on reproducible statistics, and every step can be questioned and replaced. The Super Lotto Statistical Prediction System v1.1 introduced below is built according to this philosophy. It decomposes "number selection" into several stages: hot/cold/warm statistics, gap value backfilling, zone break analysis, correlation structure identification, combination filtering, and independent back-zone analysis, while utilizing reinforcement learning to automatically adjust weights. It does not guarantee a win, but it ensures that you can clearly see the input and output of every step. **# What is this?** A system designed for the Super Lotto (Front Zone: pick 5 from 35, Back Zone: pick 2 from 12).
Read MoreDSH Simple Search: Conversation Content Search
dsh-simple-search is a minimal DSH plugin maintained by luicarstar, designed solely for searching conversation content. It features a button and a query box in the conversation header, invoking `sessions.search` with a 280ms debounce. The results display highlighted snippets and session IDs; pressing Enter opens the session, scrolls to the hit location, and triggers a flashing animation to pinpoint the position. It does not include an archive browser or settings page. Installation is enabled via `plugin_manager install_bundle` and `include:simple-search`. It has been tested against version 0.2.1-rc.5 and is released under the MIT license.
Read MoreAdvertising and Marketing Compliance Risk Assistant: Translating Advertising Law Red Lines into an Actionable Self-Audit Framework
This article introduces the "Ad & Marketing Compliance Risk Control Assistant" (@frommax/ad-compliance-assistant-v3, v1.1.0, professional, published by Fumanjaozhi Brand Strategy Consulting) on SkillHub. By internalizing 11 monographs on advertising law, Internet advertising, market regulation, and marketing risk control, and integrating 2026 regulatory updates, it provides enterprises with compliance knowledge assistance, risk checklists, and self-audit frameworks. It helps users identify red lines, recognize common pitfalls, and organize self-audits, but does not perform final content review or issue legal opinions. The article outlines its core features, configuration characteristics (a paid model requiring no API key), typical use cases, and applicable boundaries. It also accurately notes that the official page failed to be scraped this time and the installation command could not be verified.
Read MoreDeepERP Skills: Querying DeepERP MES Production Data in WorkBuddy
DeepERP Skills is a community-published WorkBuddy Skill (Coordinates: @user_93006a8e/deeperp-skills; Version 1.1.0; Category: business-ops) that queries and analyzes DeepERP business data via the DeepERP MCP tool. It covers dimensions such as MES production planning, work orders, production tasks, process reporting, work order reporting, daily production value reports, daily work order reporting, production progress, production status, source documents, project numbers, order numbers, responsible persons, responsible departments, processes, materials, output products, yield rates, defect symptoms, material deductions, warehousing, output value, cost, and profit. Users should employ this Skill when they need to query, summarize, filter, compare, or explain DeepERP MES production management, reporting management, or report management information. The tag `requires_api_key=true` indicates that an API Key must be configured before use. The article describes its positioning, scope of capabilities, prerequisites for installation and activation, typical usage scenarios, and applicable precautions.
Read Moredsh-prompt-optimizer: Rewriting Prompt Drafts with Project Context in DSH
dsh-prompt-optimizer, maintained by lucky6star9, is a prompt optimization plugin for DeepSeek Harness, categorized under SkillHub as model inference, with a brief description of prompt optimization. The public documentation states that it reads workspace structure and conventions, and upon manual user triggering (the light bulb button, the /optimize slash command, or the optimize_prompt tool), it sends the draft and project context to the model and writes the rewritten result back to the input box. The installation command is `dsh plugin --profile web add github:lucky6star9/dsh-prompt-optimizer`; the directory page shows v1.0.0. The plugin runs with DSH process privileges, and the source code and MIT license should be reviewed before installation.
Read MoreWord Teaching Assistant: Transforming Word Lists into Interactive Contextualized Learning Web Apps
Word Teaching Assistant is a web app generator for contextualized English vocabulary instruction, released by community developer user_90a020b6 on SkillHub (Education category, Version 1.0.1). By inputting a list of words (pasted directly or specified by textbook unit), it generates an interactive HTML learning application accessible on both mobile and desktop devices, comprising four sections: contextual articles, flip flashcards, an Ebbinghaus review calendar, and auto-saving notes. It is triggered by expressions such as word teaching, memorizing words, vocabulary instruction, generating articles with specific words, creating review plans, Ebbinghaus vocabulary learning, lesson preparation for English teachers, and textbook vocabulary lists. It is not suitable for pure translation, word definition lookups, grammar explanations, or exam paper generation. No additional API Key configuration is required for activation. The article introduces its features, installation access, typical usage, and applicable boundaries.
Read MoreDocument to Markdown Skill: Convert common documents to Markdown
This article introduces the SkillHub community skill "Document to Markdown" (ID: @user_75a20f40/laozhuang-b06, version 1.0.0, category: office-efficiency, no API key required). This skill allows users to convert files such as docx, pptx, xlsx, pdf, html, csv, rtf, odt, epub, and txt into Markdown or extract the body text using a single natural language command. Images and scanned PDFs are recognized via OCR and formatted as Markdown. The article outlines trigger conditions, supported input formats, the official URL, and typical usage examples. It also notes that metadata indicates zero installations and limited community verification, and that table structure and styling information may be limited in the resulting Markdown; therefore, small-sample validation is recommended before integration.
Read MoreDX-B2-06 Post-Marketing Surveillance Plan: Introduction to a Skill for Pharmacovigilance Niche Services
This article presents the DX-B2-06 post-marketing surveillance protocol on SkillHub, focusing on its intended use when physicians or pharmaceutical companies need a specialized service that provides a post-marketing surveillance protocol. Based on already retrieved materials, it compiles the Skill’s official URL, coordinates, version, source, category, publisher, rating, and download information, notes that it sits in the enterprise demand-side blind spot of B2 pharmacovigilance safety, belongs to the general professional category, and is tagged with requires_api_key. The article does not elaborate unverified feature lists, installation commands, or usage examples; it only provides the official page and a coordinate-based locating method. It indicates that this Skill is suitable for scenarios where there is a clear need for a textual framework for a post-marketing surveillance protocol, and it cannot replace registration filings, clinical trials, regulatory communications, or legal advice.
Read MoreHardware Configuration Recommendations and Verification for Weaver E10 (Non-containerized) Installation
This is an AI assistant-oriented community Skill (@user_f6995a55/e10-hardware-recommend, categorized under it-ops-security, no API Key required) that encapsulates the official hardware recommendation tables for non-containerized deployments of Weaver E10. It performs two primary functions: first, it outputs recommended server configurations based on client requirements such as concurrent user count, Xinchuang (domestic innovation) compliance, high availability needs, optional services, and architecture type (monolithic vs. microservices); second, it validates provided configurations against corresponding tiers and suggests improvements. The Skill incorporates both non-Xinchuang and Xinchuang official recommendation tables; the Xinchuang section covers Hygon C86 and Kunpeng 920 processors, and also includes recommendations and disk estimation for Elasticsearch (Weaver Search). It is applicable strictly to non-containerized deployments.
Read MoreGuide to HTML Video Rendering (Agent-oriented): Batch Rendering Videos Using Web Technology Stacks
This Skill translates Hyperframes (writing HTML to render videos, built for Agents) into a methodology for generating videos using a web technology stack. It is published by the enterprise account "Kunlun Growth" at @kunlungrowth/kunlun-html-to-video-render, version 1.0.0, categorized under design-media, and requires no API key. Targeting design and engineering teams, it uses familiar HTML/CSS/animation to batch-render dynamic content into videos, making it suitable for automation and Agent integration as an alternative to manual timeline-based production workflows. The article begins by clarifying the positioning and publisher, then lists the officially verified core capabilities: describing dynamic content with HTML/CSS/animation and rendering it into video, supporting batch rendering, and being designed for Agent invocation. Officially scraped materials did not include installation commands, configuration fields, or reproducible command-line examples, so unverified installation methods are not presented. Instead, it directs readers to the official page URL and explains that, based on the generic SKILL.md format, it can typically be placed into a project's Skill directory for Agent recognition, with specific directory names and configuration subject to official documentation. The typical workflow is described as follows: decompose content into scenes carried by HTML, use CSS to control layout and animation timing, and have the Agent batch-render the output videos. Finally, the article outlines applicable scenarios (existing web technology stacks, need for batch video production, desire for automated Agent invocation) and caveats (it is a methodology rather than a ready-made renderer; rendering capability depends on the corresponding toolchain; dependencies, configuration, and key-based billing should be verified independently before integration).
Read MoreIntroduction to the Dubbing Art Skill: Supplementing Large Models with a Professional Perspective on Film Dubbing
Voice-over Art is a SkillHub community Skill located at @user_3c6cb52e/self-ent-film-voice-over (Version 2.0.2). Categorized under design-media and sourced from the community, it is tagged with requires_api_key=false and the trigger word "voice-over." Adopting the persona of a film critic and screenwriter with a film academy background, it provides the model with a systematic framework for evaluating film voice-overs across three dimensions: dubbing art, camera language, narrative structure, and mise-en-scène. This approach addresses common issues such as voice tone mismatching characters, narrators overshadowing the performance, and flat delivery in emotional sequences. Public documentation lacks directly copyable installation commands, configuration fields, and fully reproducible examples; only official pages and interface addresses are provided. It explicitly states its limitations: this Skill offers a perspective for evaluating voice-overs rather than performing audio synthesis. It is recommended to load the Skill via its coordinate, test it with short video clips, and then integrate it.
Read MoreIntelligent Background Check Analysis: The Skill for Standardizing Investigation Protocols
This is a community Skill published on SkillHub (Version 1.0.0, Category: business-ops, Publisher: user_fa14b5bc). It requires no API Key for operation. It addresses the issues of reliance on personal experience and inconsistency in background checks. By inputting the job type, risk level, and candidate's resume, it outputs a five-step background check plan. This includes a strategy overview (4-level depth grading and reference credibility ranking), a differentiated question list for 5 job categories (Technical/Management/Sales/Product/Executive, with each question annotated with core risk dimensions, ideal response characteristics, and warning signs), guidance on deciphering 12 typical vague responses (including follow-up strategies and risk rating), automated determination of 7 "red line" items (including the appeal process for contestable items), and a standardized report template. Additionally, it includes 4 quick-reference cards, anomaly handling guidance, and a complete workflow demonstration. The article introduces what it is, core functions, installation and activation, typical usage, and applicable scenarios. Official URL: https://api.skillhub.cn/user_fa14b5bc/background-check-intelligent-analysis
Read MoreWeb Scraping MCP Server: Handing Web Page Parsing Capabilities to Models
The Web Scraper MCP Server is a community-sourced MCP Skill (identifier: @user_053d27d0/mcp-web-scraper-zx, version 1.0.1, categorized under SkillHub as "data-analysis"). It bundles 26 tools, including web fetching, JSON APIs, file downloads, HTML table parsing, link extraction, CSS selectors, XPath, HTML-to-Markdown conversion, news article extraction, and product information extraction. By parsing data first before passing it to the model, it prevents the need to load entire HTML pages into the context. The documentation details the classification of its three capability categories, the SkillHub installation entry point, and configuration fields (specifically `requires_api_key=false`, indicating no API key is needed). It also describes the invocation sequence of "Retrieve, Locate, and Transform," along with usage notes regarding its community source and an installation count of zero.
Read MoreWeCom Moments Copy Publishing System: Automating Customer Moments Content from Manual to Automatic Publishing
The WeCom Customer Moments Auto-Publishing System (@user_9203a591/wecom-moments, v1.0.1, community source, business-ops category) supports creating Moments posts, managing publication records, querying send results, copywriting template library management, and scheduled automatic publishing. It is tagged with requires_api_key=true. The page shows 100000 ratings, 871 downloads, and 0 installations. Since the retrieved materials do not include installation commands, configuration fields, or reproducible examples, the article explicitly notes that verification must be conducted on the official page. It does not fabricate content.
Read Moredsh-video-notes: Turn videos into Markdown lecture notes in DSH
dsh-video-notes is a DeepSeek Harness (DSH) plugin maintained by lucky-Luor, which registers the upstream bilibili-douyin-notes skill into DSH. It supports subtitle extraction, local ASR, key screenshots, and note templates, works without login, and does not call any paid APIs. Installation requires Node.js ≥20 and Python ≥3.10; install it via the official `plugin add` command and install the skill dependencies with pip. To use it, simply provide a video link in the Agent conversation to generate .md notes; the config.json file is located in the skill directory within node_modules, and should be backed up before upgrading.
Read MoreTypography Pairing: A Design-Media Skill Focused on Font Matching
Font pairing (self-ent-design-font-pairing) is a community-design Skill on SkillHub, version 2.0.2, categorized under design-media. It requires no API key, has a rating of 100000 and has been downloaded 884 times. Positioned for senior visual designers, it offers guidance around font pairing: hierarchy from headings to body text, combinations for mixed Chinese-English text, and pairing directions selected according to design styles, with the trigger word font-pairing. This article introduces its positioning, capability boundaries, installation method and applicable scenarios; the official page was unreachable when verifying, and the installation command and examples could not be confirmed. Please defer to the official page.
Read Moredsh-session-cleanup: Permanently Delete Session Data in DSH
DSH's built-in archiving only hides sidebar sessions; the event log `session.v4.jsonl.zstd`, projection cache, and `workspace.json` registrations remain on disk. `dsh-session-cleanup` (LucasCaleb-ggg, MIT, 1.0.0) provides a "Session cleanup" panel and a sidebar "Delete permanently..." option. It deletes the three aforementioned data locations and syncs the `workspaceRegistry` via the Host API. It supports batch deletion, unregistered session groups, and workspace registration removal; it rejects running sessions and does not touch attachments. Installation: `dsh plugin add https://github.com/LucasCaleb-ggg/dsh-session-cleanup`. Requires DSH 0.2.x Web GUI.
Read More