DSH Orchestrator: Building Verifiable and Recoverable Pipelines from Confirmed Intentions
DSH Orchestrator (dsh-orchestrator) is a community workflow plugin that organizes human-validated Intents into observable, verifiable, and recoverable pipelines. It facilitates manual gates and reconciliation within Harness sessions via tools such as orch_create, while orch view provides a local read-only dashboard and scheduling interface. Requires Node 22+; installation involves using npm pack followed by dsh plugin add tgz, and configuring the workspace in cordis.patch.yml. Currently in 0.1.0-alpha.1 development preview, licensed under MIT, with verified compatibility documented for DSH 0.2.1-rc.3.
Read MoreFull Skill Set for WeChat Channels/Official Accounts Content Stream: Full Lifecycle Skill for WeChat Ecosystem Content
This article introduces the community Skill "WeChat Official Account/Video Account Content Flow Complete Skill Set Expert". This skill covers the entire workflow for WeChat Official Account articles and Video Account short videos and live streams, spanning from topic planning, content creation, layout design, and short video scripting to dual-account synergy and private domain conversion. It is designed for teams that need to systematically manage content within the WeChat ecosystem. The article summarizes the verified capability scope, trigger words, community sources, and version information. It also notes that the official page lacks standard installation commands and that users must verify the SKILL.md loading mechanism of their chosen Agent prior to integration. No unverified installation steps or user feedback are included.
Read MoreOPC Side Business Report: Automated Aggregation of Multi-platform Daily Operations
# Introduction People running side hustles typically do not rely on a single platform. With scheduled check-ins on the Qinglong Panel, order settlements on Xiapin, selling second-hand items on Xianyu, and health check-ins, their earnings and tasks are scattered across several systems. Compiling this data manually every day requires opening individual pages to check records, which is prone to omissions and makes it difficult to ensure data consistency. The OPC Side Hustle Report Skill introduced below is designed for this scenario: it aggregates these various types of data into one place, outputs them in a daily report format, and supports both Markdown and WeChat push notifications. ## What is this? OPC Side Hustle Report is an operational daily report Skill designed for individual side hustlers. Its coordinates are @user_ccd82275/side-income-daily-report, version 1.0.0, sourced from the community, and categorized under office.
Read Moredsh-works-gallery: Browse workspace images and videos in the DSH right sidebar
**dsh-works-gallery** is a DeepSeek Harness client plugin that displays images and videos scanned from the workspace root directory as a thumbnail wall in the right sidebar, grouped by Today/Yesterday/date with the most recent items shown first. Features include filtering, filename search, fullscreen preview (with Range seek support for videos), reveal in file manager, copy path, and optional batch deletion to the system trash. Default scan root is `~/Documents/deepseek-harness`; multiple roots can be configured via `DSH_WORKS_GALLERY_ROOTS`. Install via the plugin marketplace by searching for "作品库/works gallery" or manually adding to profile bundles. Requires Node.js >= 20 and the DSH right sidebar. MIT licensed. Maintainer: 822376583-hub.
Read MoreIPO Review Key Points Assistant: A-Share IPO Issue Inquiry Skill
The IPO Review Focus Assistant is a review research skill designed for A-share IPO projects. It is based on the IMA "Practical IPO Case Library (continuously updated)" and covers issuance conditions, board positioning, information disclosure, and review practices for the Main Board, STAR Market, ChiNext, and BSE. Its trigger words include IPO, review focus, issuance conditions, and board positioning. It is published by user_44747f29, sourced from the community, categorized as professional, and is version 2.0.6 (requires_api_key=false). This article introduces its positioning, verified functionalities, activation methods, sample queries, applicable scenarios, limitations, and provides the official access link.
Read MoreRemotion Video Toolkit: Supplementing Remotion Projects with Engineering Capabilities
The Remotion Video Toolkit is a Skill built on the universal SKILL.md format, published by RealQ. Tagged as enterprise and classified under design-media, it is at version 1.0.0 and requires no API Key. Positioned as a toolkit for Remotion + React programmatic video creation, it covers nine capability categories: animation, time control, rendering (CLI, Node.js, Lambda, Cloud Run), subtitles, 3D, charts, text effects, transitions, and media processing. It is ideal for reference when writing Remotion code, building video generation pipelines, or creating data-driven video templates. The article candidly notes that the official page was inaccessible during this review, and that installation commands and configuration options should be referred to the official page.
Read MoreAI Contract Review Assistant: Leave Contract Risk Clause Scanning to Skill
The AI Contract Review Assistant (@org-2ulb7xpt/ai-contract-reviewer1) is a SKILL.md-formatted Skill published by Shuce Network, classified under the "professional" category with an "enterprise" source, currently at version 1.0.1. It targets contract pre-screening, a high-frequency, repetitive task: after uploading a contract file, it automatically identifies risky clauses, checks compliance, and provides revision suggestions, with pay-per-use pricing and no API Key required. This article introduces its positioning, verified core features, prerequisites for activation (a runtime supporting SKILL.md and completed paid authorization), the typical usage workflow, as well as applicable scenarios and limitations—including per-use billing costs, the fact that AI results cannot substitute for legal counsel, and limited community support given zero installations. No installation command is provided in official materials; the article therefore states that users should refer to the official page for details.
Read Moredsh-pet-gujiguji: DSH Codex-style Desktop Pet GujiGuji
dsh-pet-gujiguji (GujiGuji) is a Windows desktop pet for DeepSeek Harness created by 7AAAAAH. It features an independent floating window and does not modify the internal dsh pet. It synchronizes states (running, waiting, failed, idle) via `desktop/state.json` and displays Codex-style task bubbles. Features include drag-and-drop throwing, idle and click gestures, a slight bounce upon successful dsh turns, and extensive settings. It can be installed via the plugin marketplace or `dsh plugin --profile desktop add dsh-pet-gujiguji`, with optional Pillow support and exe packaging. Licensed under MIT, current version 0.8.6.
Read MoreCRM Operations Expert: End-to-End CRM Skill for Sales and E-commerce
Introducing the Agent Skill "CRM Operations Expert" (@user_3ef5c419/crm-full-chain, version 1.0.2, source: community, category: business-ops, tag: requires_api_key=false). It covers the complete closed-loop from customer acquisition, lead conversion, sales funnel, contract management, customer service, to membership systems. Tailored for platforms including Salesforce, HubSpot, as well as domestic platforms such as Taobao/Tmall, JD.com, Pinduoduo, and Douyin E-commerce, it provides capabilities including customer segmentation, the RFM model, membership tiers and benefits, point systems, repeat purchase rate analysis, churn warnings, opportunity management, CRM reporting, and KPI dashboards. This article outlines the core features and typical use cases based on the official description. It also truthfully notes that the official URL was inaccessible and installation commands could not be verified at the time of writing, as well as an inconsistency between lead scoring and profile scoring. No data or case studies have been fabricated.
Read MoreLi Hai Speaks! High School Humanities Reading and Analysis: Question Analysis Teaching Skills Across Five Subjects
Specialized reading and examination skills for high school, particularly for Senior Year 3 humanities, covering Politics, History, Geography, Chinese, and English. Centered on a core framework of question-keyword analysis, logical reasoning, argumentation, and critical thinking, the approach operates primarily through a four-step methodology (thorough examination, intensive reading, deep analysis, and strategic construction) and six specific drills. It delivers five types of outputs: teaching designs, specialized practice worksheets, demonstration scripts, learning diagnostics, and custom-created situational questions. Additionally, it integrates a reading corpus and a thinking toolkit across all five disciplines.
Read Moredsh-auto-retry: Automatic retry on model request failure
dsh-auto-retry is a DeepSeek Harness community plugin that intercepts model request failures on the agent/request-error chain, performing up to 4 retries with 2s/5s/10s/20s backoff intervals, rerunning only the current step while supporting abort for immediate termination. Based on the official README, this article covers installation commands, enabled/maxRetries/log configuration, the design trade-offs of full error interception, and contract compatibility with DSH 0.2.1-rc.3. It is suitable for agent development scenarios involving gateway jitter and occasional rate-limiting failures; the plugin operates with the same permissions as the host, so please review the source code and MIT license before installation.
Read MoreAnimated Image Generation: Generate 3x3 Grids from Text or Reference Images and Synthesize Local GIF/WebP
"Animated Image Generation" is a Skill released by Xinghua Zhilian Technology under the SkillHub design-media category, version 1.0.0. It integrates the two steps of "generating continuous frames" and "synthesizing animations" into a single invocation: input a text description or a reference image to generate a 3x3 grid of continuous frames, which are then synthesized locally into GIF/WebP formats. This covers scenarios such as generating animations from images, converting photos to GIFs, animating people, products, or pets, creating dynamic images from static ones (both image-to-video and text-to-video styles), and WeChat animations. The article follows a fixed structure: the introduction addresses the pain point of developers having to repeatedly build scripts for animation creation; next, it outlines the Skill's positioning, publisher, category, and version; the core functionality section lists only verified input paths (two methods), the 3x3 grid layout, and local synthesis capabilities. The installation and activation section honestly states that installation commands are not provided in the materials, offering only the official address and page metadata (pricing_type=paid, requires_api_key=false), without speculating on differences regarding Cursor, Codex CLI, or Claude Code. Typical usage follows the official description with a three-step process, noting the absence of reproducible command-line examples. The applicable scenarios and limitations section distinguishes suitable audiences, restrictions regarding payment and input quality, and notes that page snapshot data may change. The conclusion reviews the value and provides the official address. The entire text is restrained and fact-based, without fabricating cases, data, version numbers, or user feedback.
Read MoreFlight Price Monitoring and Comparison: A Skill for Structured Price Drop Monitoring Requests
Flight price monitoring and comparison is a community Skill, with coordinates @user_b95ee7e5/flight-price-track-new, version 1.0.1, category life-service, source community, requiring an API Key. It simultaneously searches real-time prices across multiple travel platforms, supporting flight search by route, precise multi-platform price comparison by specified flight number, low-price calendar to view optimal departure dates, creating price drop monitoring tasks, and direct data connections to multiple travel platforms. The Skill itself outputs structured price drop monitoring requests, with periodic reviews and price drop notifications handled by the host Agent.
Read Moredsh-session-recycle-tools: Web Sidebar Session Recycle Bin and Batch Delete
dsh-session-recycle-tools is a DSH client plugin that adds bulk delete and recycle bin entries above the Web sidebar settings row. The recycle list is stored in localStorage (dsh-session-tools.trash.v1), and sessions are hidden by wrapping uiWorkspace.isSessionVisible, ensuring no interference with the host's archivedSessionIds. It supports grouping by workspace, searching, restoring, and permanent deletion (invoking deleteSession). To install, run `dsh plugin --profile web add github:4wgtty4zm7-a11y/dsh-session-recycle-tools` and restart the Web service to enable the feature.
Read MoreXiaoyuzhou Podcast Transcription and Notes: Turning Episodes into Structured Markdown Notes
This article introduces the community Skill @user_fc05d2ec/xiaoyuzhou-podcast-notes-workbuddy (v1.0.1, knowledge-management category, requires_api_key=false). The workflow is as follows: parse a Xiaoyuzhou single-episode link to obtain the direct audio URL, shownotes, and chapters; when there is no official transcript, use local FunASR to perform transcription (paraformer for recognition + cam++ for speaker diarization, models accessed via direct domestic connection to ModelScope); determine the content type (finance/technology & business/person interviews/knowledge popularization/general); generate notes using WorkBuddy's built-in large model according to the corresponding template; output .md note files and verbatim transcripts. The finance template includes an industry chain logic chain, data tables, supply-demand analysis, and trend judgment. Trigger methods, applicable scenarios, and offline precautions are all based on the already fetched materials.
Read MoreSystematic Thinking Analysis: A Skill for Deconstructing Any Topic into Four Dimensions
Systemic Thinking Analysis (@user_52a81e7a/system-thinking) is a community Skill on SkillHub, version 1.0.4, categorized under knowledge-management, and requires no API Key. Given any topic, it deconstructs it across four fixed dimensions: constituent elements, real-life cases, element relationships and feedback loops, and overall purpose, thereby outputting the underlying structure and operational logic of the topic. It is ideal for establishing a structural framework when learning unfamiliar technical concepts, evaluating products or organizations, or analyzing social phenomena. It is triggered by keywords such as "系统思维分析" (Systemic Thinking Analysis), "system-thinking", "用系统的角度看 XX" (Look at XX from a systemic perspective), and "帮我系统性地拆解 XX" (Help me systematically deconstruct XX).
Read Moredsh-arc-hub: Connecting ArcDream Digital Hub to DeepSeek Harness
dsh-arc-hub is a DSH Web plugin released by maintainer 499596849-jpg. It bridges to the local ArcDream Hub (127.0.0.1:8799), providing Host tools such as arc_hub_status and arc_hub_search, along with the "Arc Dream Hub" settings page and a hub circle in the bottom right corner. To enable it, users must first install arcdream-hub and execute the official `dsh plugin --profile web add` command, followed by restarting DSH. Data remains local, persona cards are encrypted with AES-256-GCM, and the project is MIT licensed.
Read Moredsh-copy-image: One-click copy images from DSH conversations to the clipboard
## The DSH built-in image lightbox only has a close button, and images cannot be directly copied. `dsh-copy-image` is published by the community maintainer 447662. It opens a full-screen viewer when an image is clicked, supporting PNG copying, saving, 1×–8× zooming, and page navigation within the same message. On the desktop, it writes to the clipboard through the host process, bypassing the clipboard permission restrictions of the Electron rendering process. To install: `dsh plugin --profile desktop add github:447662/dsh-copy-image`. A full restart of DSH is required after installation. It is licensed under MIT, with a core requirement of >= 0.1.7-rc.1.
Read MorePyWenCai: Hithink RoyalFlush Wencai Natural Language Data Query Skill
When conducting quantitative research or developing market data tools, fetching data for A-shares, indices, funds, Hong Kong and US stocks, and convertible bonds typically requires finding data sources, writing interfaces, handling authentication, and aligning fields. Tonghuashun iWencai inherently supports direct market querying via a single Chinese sentence; however, integrating it into Agent workflows requires additional encapsulation. PyWenCai encapsulates this functionality into a Skill: in Agent environments supporting SKILL.md, users can retrieve corresponding market data by asking questions in natural language Chinese. # What is this PyWenCai (Tonghuashun iWencai Data Query) is an Agent-oriented Skill based on the universal SKILL.md format. It leverages the natural language query capabilities of Tonghuashun iWencai to obtain market data for A-shares, indices, funds, Hong Kong and US stocks, convertible bonds, and more.
Read MoreKnowledge Sports | Sports Data Analytics: Football and Basketball Data Query and Pre-match Analysis Skill
Zhi Sports | Sports Data Analysis, published by Zhi Sports, is classified as an enterprise source and professional category. The current version is 0.1.25, marked with `requires_api_key=false`. It covers competitions such as the J-League, K-League, Premier League, La Liga, Serie A, Bundesliga, Ligue 1, Champions League, Chinese Super League, NBA, CBA, and the World Cup. It provides query functions for schedules, scores, standings, as well as team and player data. By integrating recent form, historical head-to-head records, injury/suspension lineups, exclusive intelligence, and relevant indices, it offers pre-match analysis, live in-game score interpretation, and post-match reviews. It is applicable to matches involving Sports Lottery Football, Sports Lottery Basketball, and Win/Loss Lottery. The article introduces its features, access methods, typical natural language usage, applicable scenarios, and precautions.
Read MoreDSH-layers-hippocampus: DSH Layered Long-Term Memory and Cyber-Hippocampus
382quaii's memory plugin for DeepSeek Harness, dsh-layers-hippocampus (v1.5.6, MIT license, zero runtime dependencies), utilizes an append-only event log as the single source of truth, supplemented by rebuildable indexes, journaling, L2.5 trend isolation, offline graphs, and a web panel. Based on the official README, this article explains tools such as memory_remember/recall/forget, two-stage forgetting, and the deep_raw black box, as well as installation methods (pnpm / plugin_manager), large-tag governance, and usage considerations, to facilitate the implementation of long-term memory for DSH developers.
Read MoreForeign Trade·Inquiry Intelligent Quotation: Breaking Inquiry Quotation into a Set of Reusable Steps
A Skill designed for handling quotations and inquiries in foreign trade scenarios (Coordinate: @user_ad389c20/ymxpznbj, Version 1.0.0, Source: community, Category: business-ops, Publisher: user_ad389c20). Its core capabilities include calculations for EXW/FOB/CFR/CIF/DDP Incoterms, exchange rate conversion quotations, export tax rebate impact calculations, overseas inquiry handling, cross-cultural business communication and quotation style adaptation, as well as generating email drafts based on quotation results. The tag `requires_api_key=false` indicates that no additional key is needed for activation. Since the public page does not provide directly copyable CLI installation commands nor list configuration file fields, this article only provides a general activation process based on SKILL.md, and explicitly excludes three scenarios: domestic trade quotations, simple exchange rate queries, and pure logistics inquiries.
Read MoreIntelligent Generation of Training Courseware: A Shortcut from Manuals to Medical Training PPTs
Baitong AI Service Platform has released the “Training Courseware Intelligent Generation” Skill on SkillHub at @baitongai/training-courseware-intelligent-generation-skill (version 1.0.0, category: education). It addresses the problems of inconsistent messaging and scattered resources in pharmaceutical product training materials. By taking product package inserts and clinical literature as inputs, it automatically generates structured pharmaceutical training PowerPoint slides, producing a first draft of the courseware. Trigger terms cover “training courseware,” “generate courseware,” “training PPT,” “product training,” “pharma courseware,” “courseware generation,” “training materials,” “sales training PPT,” and “medical training PPT,” with an intended role in training. The tag requires_api_key=false indicates that no separate API key configuration is required. The article introduces the Skill’s positioning, core capabilities, installation and activation, a typical three-step workflow (prepare materials, initiate with trigger terms, and conduct manual review of the draft), as well as applicable scenarios and precautions, emphasizing that generated content must be verified item by item against the package insert and literature. The official installation command was not provided in verifiable sources; the article states this transparently and advises referring to the official address for details.
Read Moredsh-session-delete: Adding "Delete Conversation + Recycle Bin" to DeepSeek Harness
DeepSeek Harness natively prioritizes archiving over recoverable deletion capabilities. The `dsh-session-delete` (`dsh-plugin-session-delete` v1.1.1) plugin adds a red delete option to the conversation menu and provides a trash bin entry in the sidebar, supporting preview, restore, and permanent deletion. Data is stored in `~/.dsh/trash/session-delete/`, separate from the plugin directory. Implemented according to the 0.2.1-rc.3 contract, installation via `install.ps1` or `install-plugin.cmd` is recommended. Maintainer: 3542967095, MIT License.
Read Moredsh-coding-discipline: Injecting Persistent Coding Discipline into DeepSeek Harness
dsh-coding-discipline is a plugin released by 339006900 for DeepSeek Harness. It introduces no new tools and performs no secondary model tuning; instead, it injects coding discipline (nine rules such as the verification closed-loop, constraint ledger, minimal diff, and hypothesis-first debugging) into each request step, with optional loading of the SKILL.md long-form protocol. It supports installation via npm: `dsh plugin --profile desktop add dsh-coding-discipline`, and can be configured in cordis.patch.yml with options such as compact and dropRules. It provides metrics including the change-verification closure rate based on logs in ~/.dsh/sessions, along with a /coding-discipline dashboard, and compares behavioral changes before and after installation by checking whether each turn contains coding_discipline. Released under MIT 0.1.0; DSH must be restarted for it to take effect.
Read More