Project Planner: A Skill for Translating Vague Goals into Actionable Project Plans
Project Planner is a community-maintained Agent Skill (@user_8ee65e73/pr-project, version 1.2.4, category office-efficiency). It turns vague objectives into executable project plans and delivers five types of outputs: project plan, WBS breakdown, milestones, risk register, and schedule. It is triggered by keywords such as “项目计划,” “立项,” “排期,” “任务分解,” “WBS,” and “里程碑.” The official documentation also specifies a forced routing behavior: when these terms are detected, or when users directly paste source text, links, or screenshots, the skill must be invoked immediately without first asking clarifying questions, and without relying on generic capabilities to handle the task. It does not cover coding or technical implementation details. This article introduces its positioning, core capabilities, invocation methods, typical use cases, and applicability boundaries, and provides the official page and API URL.
Read Moredsh-bubble: Play a "glug" bubble notification sound upon task completion
dsh-bubble is a DeepSeek Harness client plugin maintained by cccchensy. It plays a program-synthesized "bubble" notification sound lasting approximately 1.05 seconds when a task round ends and output stops. It listens to api-session/status and the session list, ignoring sub-agent sessions, with a 700 ms deduplication window. The WAV audio is embedded, requiring no additional network requests. The installation command is `dsh plugin --profile <profile> add github:cccchensy/dsh-bubble`; after installation, a hard refresh is required, and activation must be confirmed on the Plugins page. It is released under the MIT license and tested on the Web interface of DSH Desktop 0.2.1-rc.3.
Read Moredsh-ugreen-nas: Integrating UGREEN NAS into the DSH Toolchain with Access Tokens
dsh-ugreen-nas is a DeepSeek Harness plugin maintained by BuvkB that connects to UGREEN NAS via UGOS Pro OpenCLI and HTTPS access tokens, without relying on SSH or master account credentials. The plugin provides 59 tools corresponding to 70 permissions across 8 major categories of the control panel token; tools for ungranted categories are not registered, and mode=readonly can expose only read-only capabilities. The installation command is `dsh plugin --profile web add dsh-ugreen-nas`; after configuring endpoint, accessToken, etc., and testing the connection, you can manage files, Docker, downloads, applications, storage, tasks, and global search within conversations.
Read MoreLast30Days Skill Tracker: A Skill for Logging Daily Learning and Generating Review Reports
A personal learning review skill under the knowledge-management category on SkillHub, located at @user_922b1001/last30days-skill-wrap, community-sourced, version 1.0.0, with the official tag requires_api_key=false. The official description claims it can automatically track daily learning progress, generate visualized reports and trend charts, and sync with note tools such as Notion, Obsidian, and Logseq; it also includes the unsubstantiated claim that it can “improve learning efficiency by more than 30%.” Upon verification, the repository it lists as its source, q15004040209-creator/last30days-skill-wrap, is actually a wrapped version of mvanhorn/last30days-skill, whose functionality is to aggregate information from Reddit, X, YouTube, Hacker News, etc. to generate research briefings. The command /last30days <topic> accepts parameters such as --emit=html and --competitors, and it depends on paid API keys including SCRAPECREATORS_API_KEY, OPENROUTER_API_KEY, and BRAVE_SEARCH_KEY, which contradicts both the description and the requires_api_key=false tag. The SkillHub page does not disclose the installation command and shows zero installs; it is advisable to verify before use.
Read Moredsh-plugin-project-mode: Project Mode Provides Persistent Coordinators and Reusable Workers for DSH
dsh-plugin-project-mode is an unofficial DSH workflow plugin (v0.1.23 preview) maintained by Burns1028. It provides a persistent Coordinator in the main session, responsible for creating or reusing Workers, reviewing returned results, and maintaining the notes.md and context index. It depends on DSH 0.2.1-rc.3 and three compatibility extensions documented in the repository. The recommended installation approach is an isolated setup via project-mode.mjs at ~/.dsh-project-mode/rc3, or pinning a git tag with `dsh plugin add` on compatible Desktop builds. It is suited for multi-Agent collaboration scenarios that require retaining coordination and project memory across tasks.
Read Moredsh-session-accordion: Collapse each round of a DSH session into a single line, keeping only the latest expansion
dsh-session-accordion is a DSH client plugin maintained by ayuanxx-max. It condenses each turn of the conversation in the rendering layer into a single line featuring a time-capsule and a question summary. By default, it only expands the most recent interaction. It does not modify app.asar. Based on the official README, this article outlines the installation commands (desktop/web profile), its operational steps (including collapse-prior and click-to-expand/collapse behaviors), its combination with dsh-session-fork-tree, as well as DOM dependencies and __acReport diagnostics. It is ideal for developers who need to quickly review the conversation skeleton in long sessions.
Read MoreQC Achievement Review and Generation Assistant: Review QC Reports and Generate Presentation PPTs
The QC Report Assistant (@user_e0723adc/qc-report-assistant) is a community skill on SkillHub, categorized under office-efficiency (version 2.0.1), requiring no API Key. Designed for QC teams in fields such as engineering construction, manufacturing, and service, it first reviews QC achievement reports to examine the rationale of topic selection, the logical integrity of each PDCA stage, the standardization of QC tool usage, the sufficiency of data support, and standardization consolidation measures. Subsequently, based on the reviewed report, it generates academic-style PPTs that conform to the structural standards for QC achievement presentation. It is triggered simply by uploading a Word/PDF report or verbally stating requirements, with no need for coding or JSON preparation. This article introduces its positioning, core functions, coordinate-based activation method, the typical review-then-generate workflow, as well as applicable scenarios and limitations.
Read Moredsh-session-search: DSH Desktop Cross-Session Message Search
dsh-session-search is a DSH desktop plugin maintained by Augl-l. It builds message-level indexes for all workspace sessions locally using SQLite FTS5. The sidebar and title bar shortcuts support Chinese single-character and keyword search, with results displayed as round rows that can be jumped to and highlighted. Installation requires obtaining the tgz file from Release; the CLI example is dsh plugin --profile desktop add. It depends on host services such as DSH desktop 0.2.1-rc.3, webServer, and sessionQuery. Indexes are stored in ~/.dsh/storages/session-search.sqlite, with no content uploaded, making it suitable for developers needing local search of historical conversations.
Read MoreAI Model Asset Management: Organizing Active Models and Asset Ledgers
Introducing the community Skill "AI Model Asset Management" (@user_8a3569c1/ai-llmops, version 1.0.3, category it-ops-security). It brings five types of assets—models, prompts, knowledge bases, evaluation sets, and agent configurations—under full lifecycle management, covering asset inventories, Model Card registration, versioning and canary rollback, go-live and decommission, data/concept/model drift monitoring, token cost governance, role-based approvals and audits. It also ships zero-dependency local tools that generate, in one click, inventory checklists, model card templates, lifecycle checklists, and cost-optimization lists. It is suitable for AI platform, engineering, operations, and finance owners, and complements policy-driven enterprise AI governance. No additional API key is required to use; official documentation does not provide installation commands, so refer to the official page before adopting.
Read Moredsh-qq-access: Hosting NapCat in DeepSeek Harness to Integrate QQ Private Messages, Group Chats, and Space Messages
dsh-qq-access is a DeepSeek Harness client plugin published by Applepinea. On Windows x64, it can host NapCat v4.18.41, automatically generate API and reporting configurations, and start/stop alongside Harness. It supports automatic replies in private chats, @-mention replies in group chats, conversion of images into model-visible content, local message archiving, and plain-text posts to QQ Zone, with a QQ integration management panel available in the left sidebar. Installation requires Node 22.14+ and Harness (refer to 0.2.1-rc.4), and the plugin is enabled via `npm ci`, `node cli.mjs setup`, and `dsh plugin add` with a local tgz file; Linux/macOS and other platforms use an external NapCat mode. Before use, please verify the source code, license, and the validation scope descriptions in the README.
Read Moredsh-homepage: DSH Sidebar Card-style Homepage
dsh-homepage is a DSH client plugin maintained by AngelosZou that provides a customizable card wall in the sidebar, aggregating entry points for sessions, projects, tools and token usage, devices and storage, core versions, Skills, and plugin security review; it supports web profile installation, external plugins registering cards, and an optional homepage model tool. Based on the official README, this article covers installation commands, environment requirements, the responsibilities of each card, and usage notes.
Read Moredsh-skill-palette: Skill palette beside the input box and call traces below answers
dsh-skill-palette (SkillMate) is a DSH client plugin maintained by AlLHHH. It provides a skill list beside the input box—classified by purpose, searchable, with recently used and favorites—click to insert /skill name. Built with around 90 Chinese descriptions, untranslated ones can be written into skill-zh.json via the default model. It optionally enhances skill utilization in the system prompt. Skill capsules of the current turn can be displayed at the end of the response or beside the input box. Installation: `dsh plugin add dsh-skill-palette` or `github:AlLHHH/dsh-skill-palette`; requires DSH >= 0.2.0-0; version 0.1.4; MIT.
Read MoreTransforming Teaching Methodology into a Reusable Framework for Material Analysis in Civil Service Exams: A Xiao Ma Ge Perspective
This article introduces the community Skill "Xiaoma Ge's Perspective Skill" on SkillHub. It crystallizes the instructional thinking framework of Xiaoma Ge from Wenchao Education for civil service exams—namely, a material-centric approach, bureaucratic mindset, and humorous expression—into a SKILL.md file that can be loaded by an Agent. It is designed for civil service essay and interview preparation guidance, instructional strategy analysis, and training in material analysis thinking. The article outlines its positioning and three core capabilities, specifies clear trigger conditions (Xiaoma Ge, Wenchao Education, civil service essay, structured interview, "material is paramount") and negative constraints (not intended for general writing guidance or analysis in unrelated fields), lists three typical usage examples (material analysis, interview responses, and instructional strategies), and highlights integration considerations such as its community source, version 1.0.0, no API Key requirement, and 0 installations.
Read Moredsh-ai-search: Three-mode AI Search Panel in DeepSeek Harness
dsh-ai-search is a DeepSeek Harness internet tool plugin maintained by 745403lkh that integrates "Search → Ad Filtering → Content Deep Reading," providing Q&A summaries and recommendations, adjustable encyclopedia entries, and a browsing feed weighted by relevance, timeliness, and popularity. It supports result language selection, search history, custom search APIs, and thinking models, and can be paired with dsh-better-sidebar. Based on the GitHub README and the SkillHub directory page, this article introduces the installation identifier, settings, the curated_search tool, and the usage of local POST endpoints, while also explaining permission and security restrictions.
Read MoreEarthwork Calculation: Offloading CASS-standard Earthwork Calculations to a Local Agent
Earthwork calculation is a general-purpose Skill released by a community (ID: @user_b2e0f551/earthwork-calc-cn), categorized as 'professional'. It requires no API Key and is currently at version 1.0.6. It strictly adheres to CASS standards by default: utilizing the CASS grid method (grid area multiplied by the average height difference of the four corners), with coordinate sequences ordered East first, where 'h' equals natural elevation minus design elevation (positive values indicate cut, negative indicate fill). It covers five calculation algorithms: site leveling, leveling elevation, earthwork balance, foundation pit sloping, and dredging/replacement, plus dual-phase earthwork, cross-section methods, and external result review. It supports four input types: scatter points, DXF files, corner point tables, and section tables. Outputs include a static Excel ledger (values only, no formulas), grid result drawings, and a 13-item self-inspection report. All calculations are performed locally without network connectivity.
Read MoreDSH Cookie Guard: DSH Desktop Login Cookie Leak Self-Check Popup
Each launch of the DSH desktop version writes a `dsh-auth-*` cookie that is not removed by default. The accumulated request headers reach the Node.js 16 KiB limit around the 72nd launch, causing "Web authentication failed." The `dsh-cookie-guard` plugin displays the cookie count, byte usage, and remaining launches via a read-only floating window on Windows. It provides color-coded warnings when approaching the limit, install commands, and instructions for an optional standalone cleanup tool. MIT 1.0.1, Maintainer: 3150481853.
Read MoreCost Engineer Question Bank · Module: Engineering Costing: Chapter-wise Exercises for "Construction Engineering Costing"
This community Skill generates practice questions for the "Construction Project Costing" subject of Cost Engineers on a chapter-by-chapter basis, covering six sections: investment estimate, design budget estimate, construction drawing budget, bill of quantities, tender quotation, and completion settlement. It does not rely on a built-in fixed question bank; instead, the AI generates questions based on the general syllabus, ensuring new questions are available with every invocation. This makes it suitable for reinforcement by chapter and targeted practice for weak areas. The official description explicitly states it is not for predicting exam questions and is unofficial; therefore, it is suitable for chapter-wise practice but not for mock exams or score estimation. As the materials do not provide specific installation commands or command-line examples, this article only provides the official SkillHub address and coordinates without fabricating installation methods. No additional API key is required for operation (requires_api_key=false); model invocation credentials are configured according to the client being used.
Read MoreAIworkin: Running Through the Entire Cross-Border E-commerce Process from 0 to 1 with AI
The Cross-border E-commerce AI Assistant (AIworkin) is a one-stop, full-process AI auxiliary tool for cross-border e-commerce (AIPDD Professional Edition) published by AIPDD. It originates from enterprise, is categorized under business-ops, version 1.0.7, coordinates at @org-bzwfrdvh/aipdd-cross-board, with the tag requires_api_key=true. The main Skill covers product selection analysis, traffic path design (SEO/ads/social media/Reddit/GEO), conversion strategy breakdown, 90-day operational execution planning, and compliance risk avoidance. Tasks are distributed through sub-skills: SEO Title Optimizer, Reddit Content Operations Expert, Cross-border E-commerce AI Customer Acquisition Video Production, Cross-border E-commerce Best-selling Product Copy Generator, Cross-border E-commerce Listing Localization, Industry Analysis Report Generator, Cross-border Ad Optimization, and AIworkin Business Networking & Intelligent Customer Acquisition Network. When users request overall workflows or next-step guidance, it enters the professional operational Workflow. Installation commands and configuration fields are not provided in the retrieved materials, so they are not fabricated in the main text; ratings and download counts in the clues are inconsistent with the retrieved materials, so the main text adopts the data from the retrieved materials.
Read Moredsh-cognition-engine: Bi-temporal Memory and Criteria Base, enabling Agents to Remember and Know When to Invalidate
dsh-cognition-engine is a memory plugin on DeepSeek Harness that provides bi-temporal memory, user-defined topics, per-turn injected guidelines (with configurable validity periods and a manual review queue), optional three-stage reflection, and long-term todos under each topic. State is stored in a local JSON file by default, with no database. Recommended installation: `dsh plugin --profile desktop add dsh-cognition-engine`; requires Node.js ≥ 22. In a new session, you can first call `cognition_context`, use `cognition_remember` / `cognition_recall` to maintain facts, and manage guideline distillation and expiration in the panel. Before installing, please review the source code and MIT license, and be aware that the plugin reads and writes local data with the permissions of the DSH process.
Read MoreFinancial Morning Briefing: A-Share Pre-Market Global Market Overview
Financial Morning Briefing (@user_b35a0bc0/caijing-zaocan, v1.8.0, SkillHub Community data-analysis category) is an A-share pre-market information Skill. It automatically outputs a snapshot of the external environment (including a Japan and Korea stock market opening module) at 08:45 every trading day. It captures global financial, military, and geopolitical news, as well as major asset fluctuations occurring between the previous day's A-share close and the current morning, assessing the positive, negative, or neutral impact on A-shares item by item and mapping relevant sector targets. At 17:00 on the same day, it automatically reviews the morning briefing's assertions and direction list to finalize the closing settlement, while iterating on the previous day's views at the beginning of the next day's briefing. Data sources are graded L1–L6; Miaoxiang serves as the sole price benchmark, ifind-mcp acts as an independent verifier, and wscn and Xinhua Finance MCP are corroborating sources. It incorporates five built-in judgment disciplines and a prohibition on reverse induction in summaries. It forms a feedback closed-loop with "Yedu Advanced Review" (zaocan_verify_ledger.json, Hard Rules A/B, and a four-step substantive integration loop). Since v1.6.2, Markdown to HTML rendering has been resident.
Read Moredsh-video-skills: Browse and One-Click Install Video Production Skills in DSH Settings
## dsh-video-skills dsh-video-skills is a DSH Web client plugin (v0.1.4, MIT) released by maintainer 1368546410qq. It displays the awesome-claude-video-skills catalog (approximately 373 entries) on the settings page, supports filtering, progress indication, and one-click installation to ~/.claude/skills or ~/.dsh/skills. It only downloads GitHub archives and copies files, without executing any repository code. It also provides three tools—video_skill_search, video_skill_install, and video_skill_uninstall—as well as the /video-skills/api/v1 endpoint. The installation command is `dsh plugin --profile <name> add dsh-video-skills`, and it must be written into `dsh.profile.bundles`. Node.js 20+ is required. Before use, please check the target repository and its license; the plugin writes to local directories using the DSH process permissions.
Read MoreWebsite Specification Skills: Multi-platform Domain Skill Pack for AI Coding Assistants
The Website Specification Skill (@user_3c6cb52e/specification-website-skill) is a community-sourced skill package categorized under dev-programming. Versioned at 1.0.0, it is positioned as a domain-specific skill module designed for AI coding assistants. It officially claims compatibility with six platforms: Claude Code, Codex, WorkBuddy, OpenClaw, QClaw, and Hermes Agent. The package is tagged with `requires_api_key=false`, indicating that no API Key configuration is necessary. As public documentation lacks official installation commands and reproducible usage examples, this article solely covers verified capabilities, installation access points (the SkillHub page and API address), enabling methods based on the standard SKILL.md format, applicable scenarios, and precautions. Notably, the source is a personal community account, and the official page displays a rating of 100,000, 1,263 downloads, and 0 installations. Due to the lack of actual usage feedback, it is recommended to review the package contents and conduct a trial deployment on a limited scale before adoption.
Read MoreDoc Toolkit: A One-Stop Processing Toolbox for Common Documents
Doc Toolkit is a document processing Skill produced by the SkillHub community, located at @user_b22a52ba/doc-toolkit, version 2.0.0, categorized under office-efficiency, and does not require an API Key. It targets common document formats such as Word, Excel, PPT, PDF, and images, providing capabilities including merging and splitting, mutual format conversion, image compression, smart cutout, PDF watermark encryption, image extraction, and batch processing, aiming to consolidate scattered document operations into a single toolbox. The article introduces the scope of features, loading methods, typical trigger scenarios, target users, and precautions, and notes that installation commands and parameter examples are missing from the scraped materials and should be based on the official page. The article also points out that the ratings and download counts provided by the clues are inconsistent with the records on the official page, and recommends using real-time page data as the reference.
Read Moredsh-workspace-classifier: Adding a Categorization Layer to the DSH Sidebar Workspace List
dsh-workspace-classifier (v0.1.1, MIT) adds a category or date grouping layer to the native workspace sidebar of DeepSeek Harness Desktop, preserving conversation rows and native interactions. It supports custom categories, two separator styles, collapsing inactive workspaces, notification dots, and one-click restoration to the native list. Requires DSH >= 0.2.0-rc.2. Install via `dsh plugin --profile desktop add dsh-workspace-classifier` and restart the Harness. Maintainer: YuyuYuu-wyt. A community plugin in the workflow category.
Read Moreworldbuilding-logic: Three-Phase Interactive Worldbuilding Skill
"worldbuilding-logic" is an Agent Skill based on the SKILL.md format, published by SkillHub community user @user_a45505fa. It is categorized under content-creation and sourced from the community (Version 1.0.1). It deconstructs worldbuilding for novels and RPGs into a three-stage interactive process: Worldbuilding Interview, Story Architecture Derivation, and World Exploration Q&A. Supporting multiple power systems—such as magic, technology, and Xianxia (cultivation)—it claims to include four rounds of self-checking involving logical deduction/rebuttal, sociological review, and economic review. This process aims to help creators produce "living worlds" that are logically consistent, naturally conflict-driven, and capable of generating derivative plots. This article covers its positioning, core features, typical usage, and usage considerations. As the official page was unavailable at the time of writing, the installation command could not be verified and is therefore omitted. The resource is marked as `requires_api_key=false`, indicating no additional API Key is needed. Please note that the ratings and download figures in the text are data points captured at the time of retrieval only.
Read More