Scheduled Automation Task Design: Decomposing Repetitive Daily Tasks into Programmable AI Task Chains
This article introduces the "Scheduled Automation Task Design" Skill on SkillHub (produced by Kunlun Growth, located at @kunlungrowth/scheduled-automation-design, version 1.0.0, category ai-agent, tag requires_api_key=false). It breaks down repetitive daily tasks, such as generating daily reports, summarizing data, and sending reports, into AI task chains that can be scheduled for automatic execution. After designing the workflow, it can be handed over to WorkBuddy to automatically run according to the plan. Built-in trigger words: scheduled automation, daily automation, automatic daily reports, periodic tasks. The article explains its positioning, core capabilities, installation and activation methods, reproducible typical use cases, and considerations such as scheduling relying on runtime, workflow granularity, data source permissions, and delivery channel permissions, and provides the official URL. During the writing process, the official page failed to be scraped and the installation command was not verified, so no fabrication was made.
Read MoreOffice Furniture Tendering Assistant: A Consolidated and Filterable Lead Pool from Scattered Procurement Notices
The Office Furniture Bid Radar is a Skill with the coordinate @org-mbvufw6y/office-furniture-bid-radar. Version 1.0.0, released by Elite Boundary, it falls under the business-ops category on SkillHub, with underlying data provided by the Shibo Bidding Data API. It integrates nationwide office furniture procurement data into a searchable interface, supporting filtering by region, usage scenario, furniture type, purchaser, keywords, budget, time range, and project phase. It covers procurement intentions, tender announcements, bid-winning results, and leads related to designated enterprises. The furniture categories cover office desks and chairs, conference tables and chairs, file cabinets, bookshelves, reception desk furniture, sofas, reception furniture, school desks and chairs, dormitory furniture, hospital furniture, laboratory furniture, and public space furniture. Procurement content covers furniture customization, supply, transportation, installation, and after-sales services. Outputs include project title, purchaser, usage scenario, furniture category, publication date, budget, quantity or delivery requirements, phase or status, matching keywords, and the original announcement link. Monitoring results highlight new additions, status changes, upcoming deadlines, and published results. The page displays a rating of 100000, 2026 downloads, 0 installations, and the tag requires_api_key=true. It does not replace users in making product certifications, environmental testing, quotations, supply commitments, legal judgments, or promises regarding bid-winning probability.
Read Moredsh-whale-pet (Big Fat Fish): DSH Desktop Pet and GalGame Dialogue Theater
Big Fat Fish (dsh-whale-pet) is the DSH community web plugin v0.6.0 maintained by Ewigkeit1203, featuring a WebGL desktop pet, display of time periods and balances for Liang Wenfeng/Liang Wengu, a GalGame theater with four scenes, and background music. To install, download dsh-whale-pet-v0.6.0.zip from Releases, unzip it, run install.bat, and refresh DSH. This is an unofficial plugin; please consult the source code and MIT/art licenses before use.
Read MoreKunlun Growth · Livestream: Deconstructing Scripts into 5 Reusable Segments
Kunlun Growth · Live Streaming (@kunlungrowth/kunlun-zhibo, v1.2.0, Source: enterprise, Category: business-ops, Publisher: Kunlun Growth) targets live streaming business growth by deconstructing livestreams into reusable scripts—opening, value, trust, conversion, and review—accompanied by supporting scripts and pacing to improve viewer retention and conversion rates. The article details its positioning, core capabilities, installation and activation methods, and typical usage scenarios. Since fetching the official page and API addresses failed this time, unverified installation commands are not provided. It adopts the standard SKILL.md format with requires_api_key=false, requiring no API Key. Note that it only covers the script framework and excludes ad placement, product selection, and data feedback.
Read MoreCapCut Quick Cut · Draft Factory: Generate CapCut Pro Drafts from Asset Lists and Subtitles
"JianYing QuickCut Draft Factory" is a community skill that takes a material list, copy, or SRT subtitles as input and generates a JianYing Pro draft project with timeline, subtitles, BGM, and fade-in/out. It supports direct export of the finished video via UI automation, but is only compatible with JianYing 6.0 and below. It only manipulates the user's local draft files and does not modify the JianYing application itself. The article explains its positioning, verified features, official address and activation method, typical workflow, as well as caveats such as missing export version and installation information.
Read MoreStock Strategy Data Retrieval & Dual-Track Verification: A-Share/ETF Market Data Skill
"Stock Strategy Data Retrieval + Dual-Track Verification" is a market data retrieval Skill (Version 1.0.1, Category: professional, Source: community, Tag: requires_api_key=false, Rating: 100000, Downloads: 1202 at the time of capture) published by user_6205de41 on SkillHub. It utilizes dual data sources—TDX and Tencent gtimg—to retrieve A-share/ETF historical daily K-lines (forward-adjusted), real-time intraday prices, and ETF premium/discount rates (IOPV). It is suitable as an alternative data retrieval channel when westock or unified market data APIs are unstable, intermittently return empty responses, or encounter errors. Typical trigger phrases include "Fetch daily K-line for [stock]," "Get real-time price," "Calculate ETF premium rate," "Switch data source due to instability," and "Stock strategy data retrieval." During the writing process, the official page failed to load repeatedly, so the installation command and client compatibility could not be verified; this has been explicitly noted in the text for accuracy. Please refer to the official address for the most up-to-date information.
Read MoreKanbanNiang: Live2D Mascot and MCP Toolchain in DSH
KanbanNiang is a DSH client plugin maintained by Evil-Wideplain: a desktop kanban-girl based on Tauri 2 and Live2D, with a built-in Streamable HTTP MCP service (11 tools, default at 127.0.0.1:7432). After installation, the process is automatically launched via link registration and cordis.patch; the tools are injected into DSH sessions as mcp__kanbanniang__*, and clients such as Claude Desktop can also be configured. The article describes the core capabilities, installation and verification steps using npm run plugin:install, and notes that one-click installation from the plugin marketplace is not currently supported, and that prerequisites such as providing your own model are required.
Read Moredsh-super-agents: Multi-Agent Delivery Pipeline in the DSH Sidebar
dsh-super-agents is a DSH workflow plugin maintained by EternalNight996 (MIT, Resource version 0.1.184). It provides a sect-style UI in the sidebar and implements staged plan approval, task DAG scheduling, quality contracts, and independent reviewer gating under the hood, with team status stored in the workspace's `.agent-teams/` directory. Install it via `dsh plugin --profile desktop add dsh-super-agents`; after restarting, the delivery loop is completed through the workflow of founding a sect, reviewing plans, approval, and library display.
Read MoreKunlun Growth · Voice Transcription: Converting Audio into Deliverable Structured Text
Meetings, customer service calls, and interview recordings require searchable, analyzable, and deliverable text; however, directly calling generic ASR APIs typically yields raw subtitles with misaligned timelines, unlabeled speakers, and homophone errors. "Kunlun Growth · Speech-to-Text" (@kunlungrowth/kunlun-speech-to-text, v1.2.0, Category: Knowledge Management) transforms audio-to-text conversion into a complete engineering pipeline: Audio Preprocessing → Transcription Engine → Speaker Diarization → Post-processing → Structuring → Quality Assurance (QA). This corresponds to six capability modules and incorporates four built-in methodologies: the "Hotword" efficiency booster, Speaker Diarization + Timestamps, Post-processing Structuring, and the Sampling QA method. The output is ready for direct delivery, and the entire workflow includes troubleshooting for pipeline bottlenecks and compliance safeguards. This article covers its positioning, core features, applicable scenarios, and important considerations.
Read MoreAsset Allocation Advisor: Skill for Generating Asset Allocation Recommendations for Private Banking Advisors
The Asset Allocation Advisor (`@user_8c473276/asset-allocation-pro`, v3.2.0) is an asset allocation Skill sourced from the Community and categorized as Professional on SkillHub. Designed for Private Banking relationship managers, it assists financial advisors in providing asset allocation recommendations to high-net-worth clients. The workflow is strictly defined in seven steps: collection and verification of client preference information, calculation of major asset class ratios via integration with `asset-allocation-ratio`, analysis and look-through of current holdings (mandatory product information query), linkage with three types of diagnosis, comprehensive analysis and optimization suggestions, product recommendation, and the generation of an "Asset Allocation Proposal." Starting from v3.1, this Skill no longer includes a built-in allocation ratio algorithm. Instead, it maps four items—risk tolerance, expected annualized portfolio return, acceptable annual drawdown level, and money market ratio—as inputs. The `asset-allocation-ratio` then outputs recommended ratios for equities, debt (fixed income), and money market assets, along with configurable ranges, five-tier comparisons, and expected indicator validation. This integrated Skill is free by default using the package-frozen snapshot; refreshing market parameters costs 0.01 RMB per use. The linkage with diagnostic tools is mandatory: ≥3 funds require `fund-diagnosis-pro`, ≥3 insurance policies require `policy-diagnosis`, and ≥3 bank wealth management products require `wealth-product-diagnosis-pro`, with conclusions quoted verbatim. Version 3.2 requires the three diagnostic tasks to be dispatched in parallel (tests show this saves nearly half the time) and requires obtaining structured Markdown summaries from each other for citation. Stripping HTML reports from counterparts is prohibited; the principle is to use HTML for human viewing and Markdown for Skill reading. For product recommendations, version 2.1 requires stating the planned product types and quantity requirements first, followed by inquiring whether the user wants to specify products, specify the recommendation scope, or search for products on sale online (default). Online search prioritizes products on sale at the user's service bank and marks the source. Data sources in v3.0 are exclusively routed through connected WorkBuddy connectors with zero Python dependencies. Fund data prioritizes Yingmi Fund (`yingmi-mcp`), followed by East Money Miaoxiang (`mx-ds-mcp`), Tonghuashun (`ifind-mcp`), and neodata. Web search acts as a fallback. External data is handled via a four-tier cascade: connected connectors → other MCP configurations → reminder to connect within WorkBuddy → non-WorkBuddy external registration. Rigid constraints include prohibitions on virtual data, skipping product queries, skipping diagnostics, and altering rule files; estimated data must be clearly marked. Available integrated Skills include `client-insight`, `wealth-planning-pro`, `asset-allocation-ratio`, and the three diagnostic Skills; if not installed, they are downloaded via `references/linked-skills-guide.md`. It is applicable for private banking client asset allocation advice, investment portfolio recommendations, holding review and optimization, and client review reports. Trigger words include "asset allocation," "asset allocation advice," "help me do asset allocation," etc.
Read Moredsh-nest: Using "Session Dependency" in DSH to query a sub-session as an agent
dsh-nest (epsilonH-eps) is a DSH workflow-class plugin: it maintains a local relations.json via a dropdown menu under Settings → Session Dependencies, allowing the parent session to wake child sessions in place through nest_list_sessions and nest_ask_session, relaying Q&A without copying conversation history or altering the sidebar session tree. It supports one child with multiple parents, returns running on timeout, and includes lineage-based cycle prevention. This article, based on the repository README, covers installation (profile symlinks and cordis.patch.yml), configuration options, typical usage patterns, and boundary conditions such as permissions and wake-up behavior, and provides links to SkillHub and GitHub.
Read MoreCounselor Workbench: A Pure Front-end Work Management System for University Counselors
The Counselor Workstation (`@user_6ff8a6ba/counselor-station`) is a pure front-end work management Skill published on SkillHub. It comprises 34 functional modules covering the entire workflow, including student management, attendance tracking, counseling sessions, academic warnings, mental health monitoring, employment tracking, scholarships and grants, and weekly reporting. Storing data in browser localStorage without the need for a backend server, it supports multi-class management, three-level privacy protection, and CSV import/export. It also provides complete source code for one-click deployment locally or to the cloud. This article describes its positioning, core features, installation methods, typical usage, and relevant considerations.
Read Moredsh-branch-worktree: Select a branch in a new session and automatically enter the Git worktree
By default, new DSH sessions start in the current workspace and current branch. Parallel multi-branch development often requires manual worktree or stash operations. dsh-branch-worktree (by elfgzp, a client plugin requiring DSH≥0.2.0 and git) adds a branch chip to the New Session line: it can automatically create wt/ branches with draft-based naming, create worktrees under .worktrees/ and register workspaces, or perform in-place checkout (refusing when the working tree is dirty). This article, based on the GitHub README, describes installation (paste repository URL), usage, and worktree/branch creation rules, and reminds users that the plugin runs with DSH process privileges and that source code and MIT license should be reviewed before installation.
Read Moredsh-prism-themes: 24 Paired Light/Dark Themes and a Theme Studio for DeepSeek Harness Web UI
dsh-prism-themes (Prism Theme Studio) provides a theme studio within Settings for the DeepSeek Harness Web interface, featuring 24 pairs of hand-crafted light/dark palettes, six moods, Palette mode, Accent character/override, favorites and "Surprise me", JSON export, and one-click Native theme. It overrides over 90 alias tokens via `ctx.theme.overrideTokens`, stores preferences in localStorage, and includes no network requests or telemetry. Installation involves registering a GitHub checkout as a bundle via the DSH `plugin_manager` or `install_bundle` (DSH >=0.2.1-rc.2). MIT License, maintained by elalamiimed.
Read Moredsh-connectome: Viewing Workspace Files and Reference Relationships via Canvas in the DSH Web UI
Connectome is a DSH bundle client plugin maintained by elalamiimed. It renders file nodes and import/link edges within the current workspace using a single canvas and d3-force in the Web UI, supporting folder regions, neighborhood highlighting, and a click-to-open right-side preview panel. Installation can be performed via the Plugins page by entering `dsh-connectome` or `github:elalamiimed/dsh-connectome`, or via the CLI command `dsh plugin --profile desktop add`. It requires DSH ≥0.2.1-rc.2, is released under the MIT License, and contains no telemetry.
Read Moredsh-tuitui-bot: Directly Chat with DeepSeek Harness in TuiTui IM
Introducing the community DSH plug-in @eayonlee/dsh-tuitui-bot: Integrates the Qihoo 360 TuiTui bot as a frontend channel for DeepSeek Harness. It supports single chats, group chats, and team channel posts, featuring in-place streaming output, interactive cards, rich media and file transmission, four levels of access control, and smart interjection detection for group chats. This guide outlines the npm/GitHub installation commands, environment variables, and settings page configuration. Please note that the plug-in runs with the permissions of the `dsh` process; review the source code before installation.
Read MoreAI-Driven Client Acquisition Video Production for Legal Professionals: From Source Material to Publishable Short Videos.
A community Skill designed for law firms, legal departments, agencies, brands, and service providers, located at @user_423c8d0a/legal-consult-intake. Versioned 1.0.0 and classified under design-media, it requires no API Key. By accepting authorized character reference images, operational environment shots, and product or service materials, it sequentially generates industry customer acquisition strategies, ensures AI avatar consistency, defines creative directions, creates storyboards, writes voiceover scripts, selects voice profiles, and designs call-to-action mechanisms. It outputs ready-to-use prompts compatible with Jimeng and Seedance. This tool is suitable for scenarios such as industry promotional videos, AI substitution for human on-camera appearances, short video lead generation, video generation from reference images, video prompt engineering, and material re-creation. The article details core features, official addresses, coordinates, and typical workflows. It also provides warnings regarding material authorization, compliance, and the availability of external services. Note that installation commands and configuration fields were not included, as the official page was unavailable and not present in the retrieved data.
Read MoreMastering IELTS Writing: Breaking Task 2 Preparation into Modular Skills
A preparation guide skill for IELTS Writing Task 2, published by community user user_ebbc273 on SkillHub (Category: Education, Version: 1.0.0, No API key required). The content encompasses strategies for four major question types, high-scoring vocabulary and sentence structures, 35 Q&As, and detailed analyses of four model essays, consolidating strategies, language materials, clarifications, and essay comparisons into a single SKILL.md file. The article outlines its positioning, four-part content structure, and activation method (the official site does not provide a standard installation command, offering only the coordinate @user_ebbc273/ielts-writing-master). It also specifies applicable scenarios and limitations: it covers only Task 2, is community-sourced, and currently shows a page rating of 100,000 with 1,189 downloads and 0 installations, indicating a small user base; it does not replace timed writing practice.
Read MoreEnterprise Bidding Inquiry: Retrieve Baidu Advertising Placement Data by Company Name
"Bidding Enterprise Query" is a data retrieval skill published on SkillHub by Xianglian Technology (@org-1qazsc77/bidding-company, categorized under data-analysis, version 1.0.2). It takes an enterprise name as input to query data regarding the enterprise's participation in Baidu bidding (SEM) promotion. It returns information such as bidding keywords, domains, search volume, and indicated prices. It is applicable to requests such as bidding enterprise queries, enterprise bidding lookups, SEM enterprise analysis, bidding promotion enterprise searches, Baidu bidding enterprise searches, and enterprise bidding data retrieval. This skill carries the `requires_api_key` tag, requiring a dedicated API Key; upon first use, it guides users to the product page to register and obtain one. This article describes its positioning, core functions, activation methods, typical usage, and applicable limitations, while noting which information has not yet been verified in public sources to prevent misuse.
Read Moredsh-pnpm-manager: Proxy and Registry Switching for pnpm under DSH
dsh-pnpm-manager is a client plugin for DeepSeek Harness maintained by EasyZoom (MIT License, version 0.1.0). It utilizes the host Config, an optional “pnpm Proxy” tab on the settings page, and a standalone CLI to exclusively modify the proxy, https-proxy, and registry entries within ~/.npmrc. It supports on/off/auto modes, mirror configurations, and connection testing. Installation is performed via the official `dsh plugin` command (supporting GitHub main branch, npm, or local paths). This tool is designed for DSH users operating behind firewalls who require local proxy configuration or frequently switch npm registries.
Read MoreAI Agent Development Consultant: Breaking Down "Creating an Agent" into Reviewable Design Cards
The AI Agent Builder Advisor (@kunlungrowth/ai-agent-builder-advisor), maintained by Kunlun Growth (Version 1.0.1, categorized under ai-agent), is not a code scaffold but a structured methodology. It deconstructs the vague enterprise concept of "building an agent" into six core elements: roles and objectives, tool/system invocation timing, memory and knowledge, human-AI collaboration boundaries, failure fallbacks, and evaluation metrics. Covering patterns such as ReAct, Plan-Execute, RAG-Agent, and multi-agent systems, along with recognition signals for agents, workflows, and rules, it ultimately produces a directly reviewable agent design card. With zero external dependencies and no API key required, this Skill is ideal for AI implementation engineers, product managers, and consultants to achieve design alignment prior to coding. The article also accurately notes that no specific installation commands have been officially released by the maintainers, and provides a generic enabling path based on the SKILL.md format.
Read Moredsh-cap-manager: Unified management of MCP, Skills, and Claude Code plugins in the DSH Web UI
dsh-cap-manager is a DeepSeek Harness client plugin (v1.2.0, MIT license) maintained by easun-zhu, requiring dsh ≥ 0.2.1-rc.2. It provides an MCP & Skills panel in the Web UI sidebar, allowing MCP management at the global/profile/project levels with real-time status and tool count display, and supports JSON import/export across multiple editors. It aggregates search, editing, zip/URL import, and backup/recovery functionality for four Skills writable root directories. It also features a built-in browser for the claude-plugins-official marketplace (315+ plugins) and supports installing Claude Code plugins. The installation command is `dsh plugin --profile web add dsh-cap-manager`; after restarting, enable it in the sidebar.
Read MoreInterior Design Rendering Prompts: 28 Styles & Floor Plan-Driven Design
Interior Design Rendering Prompts is a community-sourced Skill identified as @user_4dbce688/interior-design-skill. Version 1.0.0, categorized under design-media, and published by user_4dbce688. It focuses on interior design solutions, renovation advice, AI image generation prompts, space planning, style recommendations, and budget consultation. It covers 28 styles, supports floor plan-driven design, and produces AI image generation prompts. The skill incorporates 8 built-in strict rules, referencing standards GB 50096, GB 55031, and GB 55038. It does not trigger for pure casual conversation, architectural design, landscape architecture, or furniture purchase inquiries. The official address is https://api.skillhub.cn/user_4dbce688/interior-design-skill, and the SkillHub page is https://www.skillhub.cn/skills/user_4dbce688/interior-design-skill. Public information does not provide installation commands, directories, configuration fields, or reproducible examples; usage should be based on the official page.
Read MoreAmazon Listing Optimization Skill: Turning Deconstruction, Tracking, and A/B Testing into Reusable Processes
This Skill originates from Kunlun Growth, located at @kunlungrowth/kunlun-amazon-listing-optimizer, version 1.2.0, source: enterprise, category: business-ops, tag: requires_api_key=false. On SkillHub, it has a rating of 100,000, 1,209 downloads, and 0 installations. It is positioned as an Amazon Listing optimization manual, covering six stages: competitor analysis, keyword research and placement priority, the five elements of a title, conversion responsibility allocation across five images and A+ content, price and rating impact assessment, and post-launch A/B testing and metric tracking. It also includes four tool tables: a keyword priority matrix, a title character and placement reference table, minimum sample size and cycle requirements for A/B testing, and common compliance violation red lines. The article outlines its positioning and the core problems it solves, provides a five-step typical workflow (preparing materials, conducting competitor analysis and keyword research, rewriting the title with its five elements, planning images and A+ content, and launching with A/B testing), and describes the target audience and limitations: it does not automatically fetch backend data, ratings and download counts reflect platform-level metrics, the installation command could not be verified due to a failed scrape of the official page and API this time, and compliance red lines must be checked against the latest policies independently.
Read More16-Person Tech Value Investment Expert Group: Turning the Four Masters' Methodologies into an Investment Research Process Skill
This Skill deconstructs the value investing methodologies of four masters—Warren Buffett, Charlie Munger, Duan Yongping, and Li Lu—into an executable research workflow. It addresses two common issues: the tendency for conclusions to weaken as research materials increase, and the inability to audit errors caused by AI performing valuation and position calculations via "mental math." It features nine core mechanisms: Forced Conclusions (Pass/Fail/Gray Area + Price Range + Three-tier Position + Mirror Test), Four-Master Perspective Confrontation (debate among multiple agents after individual research), Structured Anti-Bias (A/B/C information richness ratings, 8 quick-veto red lines, and the principle of blank space), Decimal Precision Calculation (13 calculation tools, prohibiting AI mental math, and cross-validation of key data from ≥2 sources), Reproducible Output, Conflict Arbitration & Convergence, Full-link Closed Loop, Business Model Perspective, and Dual Report Delivery. It includes 32 research modes covering deep research, detailed reading of the three financial statements, industry screening, position management, decision-making mechanisms, data analysis, thinking tools, and decision journals. The assets include an `assets/workbench.html` browser-based rapid calculation panel. Dual Report Delivery requires consistency between the Word document "Investment Decision Mode Research Report" and the Markdown "Investment Research Report," verified by `verify_dual_report.py` before release. Business model analysis determines health status by completing the S-Canvas (9 sections), K-Logic (4 dimensions + 3 deepening dimensions), I-Industry Chain Positioning, and L-Cultural Proposition (4 layers). Official URL: https://api.skillhub.cn/user_7f6c97a1/shiliu-ai-berkshire Version: 2.4.0 Source: Community Category: Professional API Key: Not required SkillHub Page Rating: 100000 | Downloads: 1247 This tool is intended solely for learning research frameworks and does not constitute investment advice.
Read More