Versatile PowerPoint Expert: The Full Workflow Skill from Requirements Analysis to Final .pptx
# Introduction In the context of Agent Skills, enabling a model to directly produce deliverable PowerPoint presentations is not simple: the result is often limited to a text outline, or the generated .pptx file has disordered layouts that require manual formatting post-generation. The following introduces the **Comprehensive PPT Creation Expert**. It integrates requirement analysis, content planning, visual design, generation and delivery, and quality control into a single workflow, directly outputting downloadable .pptx finished files, an HTML preview version, and page-by-page content documentation. ## What is this? **Comprehensive PPT Creation Expert** is an Agent Skill designed for PPT presentation design and creation. Located at `@zechuan-zhuchuang/ppt-maker-expert`, it is published by Zechuan Zhichuang (泽川智创), with version `1.3.1` and source `e...`
Read Moredsh-next-sentence: Next Sentence Recommendation for DeepSeek Harness
**dsh-next-sentence (@uremysunshine/dsh-next-sentence)** provides DeepSeek Harness Desktop and dsh-tui with next-sentence recommendations generated by the current model via tab-completion. These suggestions are triggered when a turn ends and the input box is empty; they are sent as independent requests and are not written to the conversation log. Based on the official README, this article details the plugin's behavior, npm/GitHub installation commands, the cordis mounting and patch-tui patches for dsh-tui, and usage notes regarding the Esc key for closing suggestions and relevant settings switches.
Read Moredsh-code-review: Codex-style Code Review in DSH Sidebar
The community plugin @uponnoise/dsh-code-review provides a sidebar "Code Review" panel for DSH Web, benchmarked against Codex review: it collects unified diffs locally, supports uncommitted changes, baseline comparison, specified commits, and custom instructions, and can browse open PRs via the gh CLI. Reviews use the Codex rubric, streaming P0–P3 findings, and are read-only with respect to GitHub. Requires DSH 0.2.x and Git; the PR feature requires gh. Install with: dsh plugin --profile web add github:UponNoise/dsh-code-review. MIT licensed.
Read MoreCareer Planning Analysis: Multi-Platform Adaptation of Agent Response Contracts — Ensuring Consistent Representation of the Same Planning Results Across Multiple Clients
This article introduces the @qixiantech/career-path-planner on SkillHub. Powered by ARC v2.1, this Skill ensures factual consistency of career planning results across mobile, PC, pure voice, and spatial terminals, while altering presentation formats according to terminal capabilities. It supports signed fulfillment credentials and authorized cross-device retrieval without duplicate billing; meanwhile, it requires Agents to explicitly declare capabilities and safeguard credentials, as the service does not automatically identify devices. By taking inputs such as the current industry, job title, experience, goals, and self-reported skills, it generates preliminary screening and verification steps for career transitions. If the goal is unclear, it enters an exploration mode, strictly avoiding the fabrication of market salaries and industry trends. The article outlines core functions, the official page, and verifiable meta-information. Furthermore, it provides enablement strategies for Cursor, Codex CLI, and Claude Code, and highlights prerequisites to consider before integration, such as billing, credentials, and missing data issues.
Read MoreBetter Experience: Pre-send Prompt Rewriting in DSH with Persistent Sidebar Balance and Time-based Pricing
Better Experience (@unknowssq/better-experience) is a DSH community plugin maintained by UnknowSSQ for web clients. It provides a one-click draft rewrite before sending (standard/professional modes, context-aware routing, Undo), and displays account balance and peak/off-peak time-based pricing in a side panel, along with input-area draft token counts and cost estimates. Requires dsh version >=0.2.1-rc.5; while version 0.1.0 is pending npm publication, it can be installed via a local checkout. Read the source code before usage; the plugin runs with the same permissions as the host.
Read MoreElementary School English Learning Assistant (Grades 2-3): Generating Pure HTML5 Listening Pages with Web Speech API
The Primary School English Learning Assistant (Grades 2–3) is an education-category Skill published by SkillHub community author @indiv-beihai-wangliang, version 1.0.0. The page shows a rating of 100000 and 862 downloads, and it can be used without an API key. It generates a pure HTML5 English listening/read-aloud page from text under the brand name "Binhai Lao Wang Pei Ni Yi Qi Xue Yingyu", and uses the Web Speech API for TTS playback. It supports [speaker] tags, speech rate and pitch controls, and a narration language toggle (Chinese/English), making it suitable for slow-paced, zero-based learning by children in grades 2–3. The article describes its core features, how to enable it via coordinate-based retrieval, a reproducible workflow for page generation and point-to-read playback, and notes on limitations such as reliance on device voice services and that it cannot work offline.
Read MoreExcel Data Cleaner: Clean up Blank cells, Duplicates, and Messy formats such that they are ready to use
Excel Data Cleaner is an Excel data cleaning and formatting Skill published by community developer user_b0160c90. Sourced from the community, it is version 1.0.0, categorized under data-analysis, complies with the general SKILL.md format, and is tagged with requires_api_key=false, indicating it operates purely locally without an API key as described. It automatically handles common "dirty" tabular data issues such as null values, duplicates, chaotic formatting, and encoding anomalies. It covers five standardization capabilities: date unification, phone number formatting, currency unit normalization, whitespace removal, and column name standardization. It is triggered by natural language statements; users can activate it by saying "clean data," "organize table," "Excel data processing," "deduplicate," "format," or "handle dirty data" without switching tool entries or writing code. Regarding installation: No official installation commands, directory paths, or configuration fields are provided in the materials. The text faithfully explains the parts that cannot be verified, concluding only that it follows the general SKILL.md format and that the specific support scope should be subject to the client documentation. It also notes that while the page displays 115 downloads and 0 installations, the installation status and compatibility issues cannot be confirmed externally. Applicable scenarios include internal reports, event registration forms, and order details where fields are structured but text and formatting are inconsistent. Potential pitfalls to note: As a rule-based tool, it only handles in-table formatting and text issues; it does not perform cross-table verification, complete missing data, or correct field mapping. Date formatting and currency normalization merely unify the writing style and do not equate to correcting erroneous data. Given that the version is 1.0.0, the source is the community, and installation count is 0, it is advisable to validate whether it functions with your specific table structure on a desensitized sample before selection. Duplicate detection is based on whether the text remains identical after whitespace removal; rows with identical meanings but different writing styles may not be identified as duplicates. Finally, the article provides the official address to facilitate readers in verifying the current classification and data before proceeding with validation.
Read MoreTrend Analysis of Health Checkup Indicators: Transforming Historical Reports into Comparable Metric Dictionaries
The challenge of processing physical examination data lies not in analysis but in format: electronic PDFs, scanned copies, and smartphone photos all coexist, with the same indicator appearing under different names across years. This Skill (@user_65c8c185/health-report-trend-analysis, v1.2.0, published by the SkillHub community, categorized under data-analysis, and requiring no API Key) compresses parsing, indicator normalization, cross-year comparison, and report generation into a single step. The same parsing pipeline processes all three input formats, producing a unified indicator dictionary covering 114 metrics. After year-by-year comparison, it classifies results into three categories: normal, significant change, and abnormal, with separate assessments for improving or worsening trends. The trend focus defaults to the last three years but can be configured via `--focus-years`. Medical interpretations reference the Merck Manual, DXY, and Chinese clinical guidelines. Deliverables include a structured Markdown report and an offline HTML workbench, with all parsing completed locally and a de-identified dataset automatically generated. The scope of application should be clearly understood: it is designed only for physical examination and laboratory report documents, and is not suitable for stocks, bills, or contracts; the accuracy for scanned copies and photos is affected by original quality; test items outside the indicator dictionary require manual confirmation; interpretations serve as reference information and do not substitute for clinical diagnosis. This article does not include installation commands; as the official page could not be retrieved, please refer to the official address for specific activation instructions.
Read Moredsh-account-model-hub: Unified Management of Subscription Accounts and Models in DSH
UniversalCAI's dsh-account-model-hub (v0.3.2, compatible with DSH 0.2.1-rc.5) provides two views—Accounts & Quota and Model settings—within a single Web plugin. It supports OAuth/API Key login, quota and account pooling for Codex, Claude, Grok, Copilot, Antigravity, and OpenCode Go, as well as model visibility, reasoning effort, display names, and provider tool configuration. The installation command is `dsh plugin --profile web add github:UniversalCAI/dsh-account-model-hub`. The plugin runs with DSH process privileges, and tokens are stored in `~/.dsh/plugins/account-hub/auth.json`.
Read Moredsh-whale-widget-dock: Making the Whale Desktop Pet Follow the DSH Right Sidebar
dsh-whale-widget-dock is a companion add-on for dsh-whale-widget. By soft-taking over horizontal positioning, it keeps the desktop pet aligned to the Sidebar or the window’s right edge when the right panel expands, collapses, or is dragged wider, and provides a menu toggle as well as the Ctrl+Alt+L shortcut. Console configuration supports options such as gap and edgeGap. To install, run the repository’s install.ps1 or manually link it into local-plugins and register it in the bundles. A restart of DSH Desktop is required after installation. Maintained by UniMars. MIT 0.3.1. It does not modify the widget’s files; uninstalling restores the original state.
Read MoreLottery Draw Tool · Campaign Creation Assistant: Converting Everyday Draw Needs into LottoTool Configurations
This is a business-ops Skill maintained by community developer user_77dd3309 (version 1.2.7). It translates everyday lottery business requirements into LottoTool settings, covering campaign planning, creation, discovery, metrics viewing, and data export. It supports five gameplay modes: wheel, capsule (gacha), golden egg smash, 3x3 grid, and flip-flip. Configuration includes scheduled draws, instant draw, draw-when-full, prizes and number of prize tiers, win probability, participation limits, draw rules, and campaign time windows. The official page is currently unavailable, and captured materials do not provide verifiable installation commands or examples. Accordingly, this article only outlines the usage path and precautions, without fabricating commands or cases. It suits developers and operations staff who need unified management of lottery campaign configuration and data. Feature tag: requires_api_key=false.
Read Moredsh-theme-matrix-glass: Binary Digital Rain and Frosted Glass Theme Loaded at Task Runtime
The dsh theme plugin dsh-theme-matrix-glass (v1.0.1, MIT), maintained by undefineken, renders a binary Matrix digital rain (0s and 1s) on the shell overlay, activating only during running sessions. When a session stops, falling columns are immediately cleared from the canvas. It utilizes overrideTokens to overlay frosted glass panels, supporting distinct ink colors for light and dark modes, a Plugin Manager configuration card, and localStorage persistence. Installation is available via the Community Theme Market or by running `dsh plugin --profile desktop add github:undefineken/dsh-theme-matrix-glass`. It operates without shell access or network dependencies, making it suitable for developers seeking a status-oriented theme within a desktop profile.
Read MoreSKILL.md: Q&A Assistant for "Qingfeng Campus Navigation" at Henan Vocational College of Supervision
SKILL.md is a general SKILL.md-format skill published on the SkillHub community at the coordinate @user_88a922fd/rr. It is maintained by user_88a922fd, with version 1.0.0, categorized under education, and tagged with requires_api_key=false. This skill is activated when users inquire about the "Qingfeng Campus Navigation" at Henan Supervision Vocational College, covering guides for on- and off-campus life as well as preparation material packages for discipline inspection and legal certificates (e.g., Judicial Examination/Civil Service Exam). The article describes its positioning, verified capabilities, and installation via the general SKILL.md directory format. It also notes the community source and advises that the specific granularity of materials should be subject to the content within the skill after installation, ensuring no fabricated features.
Read MoreRapid Health Screening of Mechanical Vibration: ISO 20816-3 Assessment of Arbitrary-Format Vibration Signals
Mechanical Vibration Health Quick Screening (@user_09377e9a/vibration-quick-screen, v1.0.0, community source, category: data-analysis) standardizes the solution to the challenges in pre-processing rotating machinery vibration signals. Featuring a built-in ingestion layer, it automatically identifies formats such as CSV/NPY/MAT/WAV/XLSX/Parquet, sniffs metadata including sampling rates, normalizes units, and requires user confirmation for low-confidence items before proceeding. Upon confirmation, it utilizes core evaluation algorithms compliant with ISO 20816-3:2022 to output A/B/C/D zone classifications and a 0-100 health score. It is suitable for point inspections, routine patrols, and rapid on-site health screenings for any shaft-coupled rotating industrial equipment, such as motors, pumps, fans, compressors, and steam turbines. Tags: requires_api_key=false, no additional API Key required. The article outlines core capabilities, data flow, installation and activation, typical usage, applicable scenarios, and limitations.
Read Moredsh-safe-session-delete: Adding Restorable Session Deletion to DSH
DSH natively provides an archive function but no deletion entry. `dsh-safe-session-delete` offers a deletion confirmation flow via the sidebar menu, hover buttons, and "Settings → Session Management": the host side first stops the task, then moves the session JSONL into `session-trash`, cleans up projection caches and sub-sessions, and retains data for 30 days by default, allowing manual recovery. Installation requires running `dsh plugin add` from the GitHub Release v0.2.0 tarball and restarting. Compatible with DSH >=0.2.1-rc.5, MIT licensed.
Read MoreObjective and Plan Management: Breaking Down Vague Goals into Actionable, Trackable Plans
Goals and Plan Management is a business-ops skill released by Kunlun Growth to boost planning productivity. It first uses SMART to rewrite fuzzy goals into measurable definitions, then breaks them down via WBS into hierarchical work packages, and finally sets cadence while reserving buffers for unforeseen events, making plans trackable and reviewable. The article introduces its core capabilities, official activation entry point, and typical use cases, suitable for cross-team goal implementation, operational campaign planning, and R&D project scheduling. It also reminds that this skill outputs a planning methodology and still requires human confirmation on resources, dependencies, and risks.
Read More8D-CAPA-PDCA Three-in-One Integrated Workflow: Turning Customer Complaint Handling into an Automatic Pipeline
The 8D-CAPA-PDCA Integrated Workflow is a "Business-ops" skill released by the SkillHub community (Version 1.0.0, located at @user_2241deda/8d-capa-pdca-integrated-workflow). Upon inputting customer complaint information, it automatically generates 8D reports, tracks CAPA tasks, and performs standardized PDCA archiving. Data is automatically transferred and status synchronized across all three stages without requiring an API Key. It is suitable for scenarios such as full-process customer complaint handling, rectification of internal audit non-conformities, supplier issue closure, and process anomaly improvement. The installation entry is via the official address; however, public documentation does not provide specific installation commands, so users must visit the page to confirm details based on their specific tools. Pre-fetched data indicates a rating of 100000 and 835 downloads, though this conflicts with other sources; the actual page display should be considered authoritative.
Read MoreMoyu (dsh-plugin-moyu): Windows Audio Visualization and Lyrics in DSH
Tywim (TYWIM/dsh-plugin-moyu) is a community plugin for DeepSeek Harness. On Windows, it captures SMTC and WASAPI loopback audio and performs 48-band FFT analysis via a PowerShell media bridge. The data is pushed to official UI slots via SSE, providing “Normal” and “Whale Petting” modes, local and online lyrics, and customizable color schemes; it remains dormant on non-Windows platforms. Based on the README and SkillHub directory documentation, this article covers its positioning, configuration, and usage precautions.
Read Moredsh-subagent-setup: Delegate subtasks to designated subagents and models by alias
dsh-subagent-setup is a DeepSeek Harness plugin maintained by twoweeks1995 that delegates tasks to specified subagent agents and models via aliases. Configuration is stored in DSH_HOME/subagent-setup/config.json and a Chinese settings page, with no YAML writing required. It provides five tools including subagent_route, background delegation with wait convergence, a routing capsule in the session header, and a release button. Installation supports the npm package name, GitHub link, or Releases tgz. The verified environment is DSH 0.2.1-rc.3 and Node≥20. MIT licensed, version 0.2.2.
Read MoreRapid Industry Research: Delegating "Quick Industry Understanding" to an Agent Skill
The "Industry Quick Research" is a professional Skill under the community category on SkillHub, located at @user_b8af0005/industry-quick-research. Version 1.6.0, published by user_b8af0005, with a page rating of 100000 and 858 downloads. The tag requires_api_key=false indicates no additional API key is needed. The official description consists solely of "Quickly understand an industry —" without listing research dimensions, data sources, output structure, or reproducible examples. The official page failed to load during this retrieval attempt, and the provided materials neither include installation commands nor configuration fields. Consequently, the article does not contain unverified commands; it merely presents the general skills directory conventions in the SKILL.md format, clearly marked as unofficial statements. The article outlines the Skill's positioning and core functional boundaries, explains the natural language calling method, and highlights important notes such as manually verifying primary data, validating output structure with a familiar industry before first use, and the significant disparity between download counts and rating magnitude. It concludes by providing the official address.
Read MoreDoubao Q&A Skill: Browser automation to ask Doubao questions and retrieve answers
Doubao Q&A is an ai-agent Skill published by user_d8831e99 in the SkillHub community. It uses the general SKILL.md format, currently at version 1.0.0, and is located at @user_d8831e99/doubao-qa. It automates the process of asking questions and retrieving answers from Doubao AI (doubao.com) through a browser: it opens Doubao, submits a query, waits for the response, and extracts the answer. No login or API Key is required (requires_api_key=false). It supports trigger words such as "问豆包", "问豆包关于", "豆包查询", and "豆包搜索", and can be routed directly within an Agent. Typical usage includes ad-hoc queries (e.g., looking up nearby phone shops, knowledge points, or tutorials) and batch generation of knowledge base content using Doubao. Prerequisites: the running environment must be able to access doubao.com; wait times depend on how long it takes Doubao's page to generate a response; no timeout or retry configurations are provided in the documentation. The real-time metrics captured from the community page—score of 100000, 887 downloads, and 0 installs—are verified accurate. Answer quality is determined by Doubao itself; the Skill only retrieves content and does not perform any verification.
Read Moredsh-quiet-subagents: Main session silently receives sub-agent messages
dsh-quiet-subagents is a DSH workflow plugin (v0.2.1, MIT) maintained by Twens03. After entering `quiet_on` in the main session, `send_message` requests and completion notifications from direct child agents will not interrupt the current main turn. Instead, they enter the QueueDock and are delivered as a single readable bubble in the next turn, automatically closing the silent mode. This is suitable for scenarios involving long-running tasks in the main session or where subprocesses only assist with collection or scripting. Installation example: `dsh plugin --profile web add github:Twens03/dsh-quiet-subagents` Requires DSH >= 0.2.1-rc.2; options such as `autoOff` and `settledDelivery` can be configured on the plugin details page.
Read MoreLandscape Wide-Angle: Breaking Down Visual Decisions in Landscape Photography into an Actionable Process
"Landscape Wide" is a community skill on SkillHub, situated at `@user_3c6cb52e/self-ent-photo-landscape-wide` (Version 2.0.1, published by `user_3c6cb52e`, categorized under `design-media`). It requires no API Key and is activated by the trigger word `landscape-wide`. To address the common issue of AI models providing only generic advice—such as "mount a tripod, stop down the aperture, or find a foreground"—this skill focuses specifically on wide-angle landscape photography. It deconstructs the photographer's compositional decision-making process into actionable steps, emphasizing the philosophy of "seeing" before "shooting." The output is primarily dedicated to explaining visual decision-making. The article introduces the skill's positioning, verified core capabilities, reproducible usage paths based on trigger words, applicable scenarios, limitations, and provides the official link.
Read More@tuweizhi/dsh-pet: DSH Desktop Pet Enhanced Fork
Let me analyze this text. It's a description of a software fork that provides a desktop pet for DeepSeek Harness, with specific technical details about features, installation, and behavior. The text contains technical jargon, npm package names, and specific terminology about UI behavior. I need to ensure accurate translation of technical terms while maintaining the structure and flow of the original. 1. "是...的增强 fork" → "is an enhanced fork of" 2. Package names should be preserved as-is: @tuweizhi/dsh-pet, PC2005-cloud/dsh-pet 3. "透明动画桌宠" → "transparent animated desktop pet" 4. "浏览器 overlay 与桌面置顶小窗" → "browser overlay and desktop pinned window" 5. "相对上游" → "Compared to upstream/the upstream version" 6. "碎碎念全本地语句池零 token" → "idle chatter uses a fully local phrase pool with zero tokens" 7. "气泡点击收起" → "click to collapse bubbles" 8. "自动行为间隔 action" → "automatic behavior interval actions" 9. "拖拽甩抛不出视口" → "drag and throw without leaving the viewport" 10. "贴角边距只算角色本体" → "corner-edge margins only count the character body" 11. "安装" section needs to preserve command syntax exactly 12. "重开生效" → "takes effect after restart" 13. "主动对话仍走模型" → "proactive conversations still use the model" 14. "其余运行时零 LLM" → "the rest of the runtime uses zero LLM" Let me ensure the flow is natural in English while preserving technical precision. The text is quite dense with features, so I'll maintain the compressed style but make it grammatically correct in English.</think>@tuweizhi/dsh-pet is an enhanced fork of PC2005-cloud/dsh-pet v0.3.8 (npm 0.4.4), providing a transparent animated desktop pet for DeepSeek Harness, supporting browser overlays and desktop pinned small windows. Compared to upstream: idle chatter uses a fully local phrase pool with zero tokens, bubbles can be collapsed by clicking, automatic behavior interval actions, dragging and throwing doesn't leave the viewport, and corner-edge margins only count the character body itself. Installation: For desktop apps, enter @tuweizhi/dsh-pet in plugin management; for dsh web, execute `dsh plugin --profile web add @tuweizhi/dsh-pet`, takes effect after restart. Proactive conversations still use the model, while the rest of the runtime uses zero LLM.
Read MoreWang Xiaobao Real Estate Project Marketing Daily Report Generator: Aggregating Project Visits, Profiles, and Transaction Data into an HTML + PDF Daily Report
A community Skill posted by user_2d5fa379 on SkillHub. Based on the universal SKILL.md format, it calls the Wangxiaobao API to fetch three types of data: visits, customer profiles, and transactions, generating a structured HTML and PDF Marketing Daily Report. The page is tagged with `requires_api_key=true`, requiring the configuration of a Wangxiaobao API Key before use. Trigger words in conversations include Marketing Daily Report, Project Daily Report, Generate Daily Report, and daily report. Version 1.0.0, categorized under office-efficiency; at the time of scraping, it had a rating of 100,000, 847 downloads, and 0 installations. The provided documentation does not specify installation commands or configuration field names; the body text states this factually without fabrication.
Read More