DSH Stock Research: An Evidence-Based Individual Stock Research Plugin for A-Shares and US Stocks
@megapandas/dsh-stock-research is a workflow plugin for MegaPandas built on DeepSeek Harness. Driven by fixed data snapshots, it facilitates individual stock Q&A, quick/deep research, review, and re-runs. The main Agent coordinates multiple roles to complete long-short discussions and risk assessments, delivering traceable reports without executing trades. It requires Node.js 24.16.0 and the DSH 0.2.1 series, installs via `dsh plugin add @megapandas/dsh-stock-research@next`, and supports data sources such as Eastmoney and Yahoo Finance, along with Web settings.
Read Moredsh-job-ledger: Local Job Application Ledger — Deduplication, State Machines, and BOSS Reconciliation
dsh-job-ledger is a DSH web plugin maintained by Me106y, categorized as a memory plugin. It records job application events locally via events.jsonl and provides deduplication, a state machine, an evidence chain, and reconciliation capabilities with BOSS Zhipin. Ten model tools and a four-screen panel share a single submission entry point, job_ledger_apply. The installation command is dsh plugin --profile web add dsh-job-ledger, and both data and Chrome login credentials are stored locally on the machine. The README notes that full verification on real devices has not been completed, so the source code and privacy trade-offs should be reviewed before use. Licensed under MIT 0.1.0.
Read MoreHigh School English Reading and Review at BNU Taizhou Affiliated High School: Transforming "Readable but Un-answerable" into a Teachable and Diagnostic Instructional System
An English subject extension of 'dushen-thinking-teaching,' dedicated to cultivating reading and reviewing cognitive skills for high school students, particularly those in their final year. Its core is built upon four dimensions of thinking—keywords, inference, argumentation, and critical analysis—and a four-step methodology of Detailed Auditing, Intensive Reading, Deep Reflection, and Clever Construction. A comprehensive competency spectrum for the "Continuation Writing" question type has been established, featuring keyword analysis tables, three-tier conflict models, story beat sheets, and a repository of 959 sentence chunks. Supported by four sets of textbook editions, graded original works, selected foreign press materials, past exam databases, and school-based lesson examples, it delivers six categories of outputs: instructional designs, task sheets, teaching scripts, diagnostic assessments, test construction protocols, and corpus retrieval. Currently, a methodological framework skeleton exists for Reading Comprehension, while competency spectrums for Cloze, Grammar Filling, Practical Writing, and Listening remain under development.
Read MoreGuanlan Official Document Writing Expert: A Skill for Drafting, Format Lookup, and Quality Inspection
Guanlan Official Document Writing Expert (@user_9abc1b19/official-document-writer-gwxzzj, v1.0.1, community source, content-creation category, requires_api_key=false) is a general-purpose SKILL.md format Skill tailored for official document writing. It covers standard official document generation, format specification inquiry, content optimization, and quality inspection for document types such as notices, reports, letters, and meeting minutes, corresponding to four usage scenarios: drafting documents, checking formats, modifying content, and assessing quality. The article addresses the practical pain point of formatting rework. It provides directly citable facts such as identifiers and versions, usage instructions organized in the order from drafting to inspection, and notes on target users and limitations: the scope of document types requires practical confirmation, current official document standards must be checked prior to formal submission, and generated content requires manual fact-checking. As the official page scraping failed, unverified installation commands were not included. Official URL: https://api.skillhub.cn/user_9abc1b19/official-document-writer-gwxzzj.
Read Moredsh-inline-edit: Inline Editing of Messages and Attachments in DeepSeek Harness
dsh-inline-edit is a DeepSeek Harness workflow plugin maintained by maxy1817 (v0.2.5, MIT license). It allows inline editing of text and attachments next to messages via a pencil icon, with optional default settings of "add branch" or "regenerate in current conversation." It replaces the model context through the Session API without physically deleting logs. Installation requires a pre-built tgz file, using the command `dsh plugin --profile desktop add "./dsh-inline-edit-0.2.5.tgz"`; it is not yet published on npm or any marketplace. It has been verified with DSH 0.2.1-rc.3. Note that the editable scope differs between the main message and queued supplementary messages, submission should not occur while tasks are running, and regeneration does not undo already-executed tool and file operations.
Read Moredsh-company: Order-Driven Multi-Agent "Small Company" DSH Plugin
dsh-company is a workflow plugin (MIT, v0.5.0) developed by makooslee for DeepSeek Harness. It simulates a CEO and multiple employee agents within the Harness environment, where each agent possesses an independent identity, model, and skill whitelist. Upon task broadcasting, employees acknowledge tasks by replying with "[Accept]" or "[Decline]"; those who accept execute tasks in parallel, with results converging into the team timeline. The plugin supports dispatching via an admin panel, the `company_dispatch` function, the `/dsh-company` slash command, and agent-team mirroring. As of v0.5, it also features an n8n-style workflow canvas. Installation requires Node.js ≥20, utilizing the `install_bundle` function in `plugin_manager` to mount local bundles. Company name, `cwd`, and acceptance/execution timeouts can be configured in `cordis.patch.yml`.
Read MoreCourse Video Transcription and Learning Distillation Assistant: Organizing Lecture Videos into Structured Learning Artefacts
The Course Video Transcription and Learning Distillation Assistant (@user_cae83fba/course-video-learning-distiller) is a community Skill published on SkillHub, version 2.0.0, categorized under education, and maintained by user_cae83fba. It reads course webpage AI transcripts, subtitle tracks, or local audio/video files, and distills learning notes with source markers, knowledge frameworks, and executable skill packages, without requiring an API Key. The article introduces core features, acquisition methods, four typical use cases (organizing notes, reading AI transcripts, extracting subtitles, distilling skills), as well as applicable scenarios and limitations: it is not suitable for bypassing login and paywalls, copyrighted courses requiring full verbatim transcription, or non-educational content such as music and film/television materials.
Read MoreBest-Seller Recreation: Deconstruct Structure First, Then Change Selling Points
"Viral Replication" is a community Skill designed for content creation (Coordinate: @user_93c42d4c/xy-replica, Publisher: user_93c42d4c, Version: 1.0.2, Category: content-creation, Source: community). It deconstructs benchmark viral videos into structure and transferable formulas based on graded evidence, strictly avoiding speculation. Upon obtaining the actual selling points of the user's product, it generates replication scripts or AI video generation prompts. Before receiving usable selling points, it delivers only structural analysis without fabricating selling points or visualizing product appearance. Trigger phrases include "Help me break down this viral video," "How to replicate the structure of this video," "Make one for my product based on this video," "Replicate with a different product," and "Write video generation prompts for me." Tag: requires_api_key=false. Rating: 100000, Downloads: 1279, Installs: 0. No official installation command is provided; users must check the SkillHub page for details. It is suitable for content teams with existing selling points and clear benchmark materials.
Read MoreDeepGarden IPO Research: Turning the Prospectus into an Actionable Subscription Decision
DeepGarden IPO Research is a Skill designed for in-depth research on A-share new stocks and companies. Located at @user_159a690e/deepgarden-ipo-research, version 1.0.2, it is maintained by community publisher user_159a690e and categorized as professional. It codifies a "Five-Dimensional Stereoscopic Evaluation System" into the research workflow: Market Prospects, Competitive Landscape, Operational Characteristics, Financial Characteristics, and Enterprise Quality, with a core question checklist provided for each dimension. Outputs include a structured scorecard (scoring 1–10, certainty level, a one-sentence conclusion, and mandatory ⚠️ risk signals), data traceability (primarily prospectuses and annual report financial notes, supplemented by financial data skills, Qichacha, and WebSearch), as well as subscription recommendations (Aggressive/Cautious/Give up) and confidence intervals for key assumptions, ultimately generating Obsidian notes with scorecards and disclaimers. Part of the DeepGarden investment research series, it can be installed independently or combined with deepgarden-earnings-tracking and deepgarden-report-deep-read as needed to form a closed loop of pre-subscription research, post-listing tracking, and research report verification. To enable it, say "Use the DeepGarden IPO Research framework to analyze <Stock Name/Code>" in the conversation. The resource is marked as requires_api_key=false. As official resources do not list installation commands or configuration fields, this article does not provide speculative installation steps.
Read Moredsh-plugin-muyu: Electronic Wooden Fish, Sutra Chanting, and Token Merit Ledger for DSH Workspaces
dsh-plugin-muyu is a DeepSeek Harness plugin (v0.2.0, MIT license) maintained by Makesupper. It provides a 3D floating wooden fish widget in the workspace, streams 8 sutras line by line, and offers Cantonese/Mandarin chanting audio. The server side monitors token usage per session/event and, by default, knocks the wooden fish once for every 10,000 tokens while recording merit. Together with six tools such as muyu_knock, as well as manual knocks on the widget and automatic practice, entries are written to the same merit ledger and atomically persisted to ~/.dsh/muyu-data/. Installation is recommended via `dsh plugin --profile web add dsh-plugin-muyu` or by adding the package name on the desktop plugin page.
Read MoreDeepSeek Harness SSH Toolkit: Running DSH Sessions on Remote Machines via SSH
The DeepSeek Harness SSH Toolkit (dsh-ssh-toolkit) is a DSH 0.2 client plugin maintained by MaiwulanjiangMaiming. It leverages SSH to select a directory on a remote Linux/macOS system as the workspace, with seven categories of tools—including bash, read, write, and edit—routed to the remote host for execution based on the session's current working directory (cwd). It supports importing hosts from ~/.ssh/config, ProxyJump bastion hosts, SFTP file trees and ssh -tt terminals, remote background tasks, and sandbox alignment. A full restart of DSH is required after installation, with configuration performed via Settings → SSH Connections. Passwords are stored in plaintext in local storage; please review the source code and MIT license before installation.
Read MoreFlowchart and Architecture Diagram Generation: Directly Produce Renderable Mermaid / PlantUML Code
This article introduces the Skill "Diagram and Architecture Diagram Generation" (@kunlungrowth/diagram-as-code, v1.0.0, published by Kunlun Growth, category: dev-programming). It generates Mermaid / PlantUML code with a single click based on natural language descriptions of business processes or system architectures, which can be rendered as flowcharts, architecture diagrams, sequence diagrams, and ER diagrams. The trigger words are "flowchart", "architecture diagram", "Mermaid", and "sequence diagram", and activation requires no API key. The article outlines its core features, entry points and coordinates, typical usage, and applicable scenarios. It also notes that rendering capabilities depend on the target environment, description quality affects the output, and limited public documentation necessitates prior verification.
Read Moredsh-openrouter-sync-models: Keeping DSH's OpenRouter model list in sync with the official directory
Introducing the community plugin dsh-openrouter-sync-models (m1khal3v, MIT, 0.1.7): This plugin synchronizes the `api` and `models` for the `llm-pi-ai` router via the OpenRouter authenticated API, triggering on startup, on a schedule, and when keys change. It supports `@preset` and configurable filtering. The installation command is `dsh plugin --profile <profile> add dsh-openrouter-sync-models`. It also outlines the routing, credentials, and optional `config` fields.
Read MorePro Speech Writing: Turns occasion, audience and duration into a ready-to-deliver script
Speech Draft Pro is a content-creation Skill released by Kunlun Growth, version 1.0.0, located at @kunlungrowth/speech-draft-pro, and does not require an API key. Given four inputs—occasion, audience, core message, and duration—it produces a complete speech draft or speaking outline, including an opening hook, main structure, memorable lines, interaction design, and closing. It is suitable for managers, trainers, and hosts in public occasions such as annual meetings, press conferences, weddings, toasts, and training sessions. The article outlines core capabilities, typical prompt examples, applicable scenarios, and precautions, and truthfully notes that the official page fetch failed and the installation command could not be verified.
Read MoreData Packet Capture Code: Browser Interface Packet Capture Skill Based on Python + Playwright
"Packet Capture Code" is a community-sourced Agent Skill, released and maintained by user_7e77ad98, version 1.0.0, categorized under data-analysis. It uses Python + Playwright to capture XHR and Fetch requests and responses emitted by pages within a real browser context, supports mocking or modifying browser requests, and can generate HAR or CSV reports. It is suitable for web interface scraping, API analysis, front-end interface reverse engineering, and dynamic data collection, but explicitly not for capturing raw TCP/UDP network-layer packets (Wireshark or scapy should be used), nor for mobile native App traffic (mitmproxy with frida should be used).
Read Moredsh-wsl-folder: Adding WSL Directory as Workspace in DSH
dsh-wsl-folder is a DSH client plugin (v1.1.1) maintained by LyuYuxin, targeting DSH Desktop and Web on Windows. It enables WSL directory browsing via "Add Workspace," allowing users to register Linux paths as workspaces. Sidebar terminals are launched via `wsl.exe` within the corresponding distribution, and the plugin fixes agent write issues on `\\wsl.localhost` as well as certain UNC path problems. It requires DSH >= 0.2.1-rc.3 and WSL; a full restart of DSH is required after installation. Behavior can be configured through four options: terminal, write, paths, and remote.
Read MoreQimen Dunjia: Chart Setup (Cultural Experience Edition): Real-time Arrangement of Time-based Qimen Chart Layouts into Plain Language
This is a universal Skill for generating and arranging Qimen Dunjia charts. Published by user_2a1415c1, it is Version 1.0.8, sourced from the community and categorized under life-service. No API key is required. It performs calendar calculations based on traditional "Time School" Qimen rules: determining Yin and Yang Dun (Escape) and bureau numbers via solar terms, identifying Zhi Fu and Zhi Shi via the Three Yuan system, and arranging Spirits, Stars, and Gates across the Nine Palaces. It instantly organizes the Four Pillars, solar terms, Dun bureau, Xun Shou (Cycle Head), the palace locations of Zhi Fu and Zhi Shi, and elements of the Nine-Palace chart, while providing vernacular explanations for each term and its source in ancient texts. It is entirely free, requiring no software installation, downloads, internet connection, or user accounts, and it does not collect personal information. It does not make judgments on auspiciousness or inauspiciousness, provide conclusions on choosing times or dates, or offer suggestions for changing luck or resolving misfortune. Trigger words cover Qimen Dunjia, Qimen, Dunjia, Qi Ju, Pai Pan, Nine-Palace Chart, Eight Gates and Nine Stars, Time School Qimen, Numerology, Zhouyi, Yijing, Sinology Culture, Traditional Culture; it is also triggered by questions like "which day is good" or "can this matter succeed," and responds from the perspective of organizing traditional cultural data. As different schools have different arrangement methods, this is for reference only and does not constitute a basis for decision-making.
Read MoreGEO (Generative Engine Optimization) Expert: Making Content Easier to be Cited by AI Search
GEO (Generative Engine Optimization) Expert is a community-published Agent Skill, version 1.0.1, categorized under content-creation and requires no API Key. It helps increase the citation rate and visibility of content, brands, and websites in AI generative search engines such as ChatGPT, Perplexity, Google AI Overviews, Gemini, and Claude, targeting five core scenarios: content GEO optimization, brand AI visibility enhancement, GEO audit and analysis, GEO vs SEO strategy comparison, and AI citation rate diagnosis. Trigger keywords include GEO, generative engine optimization, AI search optimization, AI answer visibility, AI citation rate, and "make AI cite me." Published by community developer user_90965760 and distributed via SkillHub.
Read Moredsh-spend-meter: View Balance, Last Turn, and Today's Spending in Real-Time within DSH Conversations
dsh-spend-meter (package: dsh-spend-panel) is a DSH client plugin maintained by lx950 that provides a draggable spend panel in desktop sessions. Data is sourced primarily from the DeepSeek platform's total_costs billing counter (1e-8 CNY), displaying balance, last-round charge, today's total, and the ongoing round; peak/off-peak time windows and failure paths are explained in the UI. Unsettled or pre-install usage is covered by a fallback via a token price list and transcript backfill. Requires DSH ≥0.2.0-rc.2 and the desktop profile. Install from GitHub: dsh plugin --profile <name> add github:lx950/dsh-spend-meter. MIT-licensed, no telemetry.
Read MoreChatDistill - Information Distillation Tool: Distill Chat Records into 26-Scheme Prompt Blueprints
ChatDistill is an information distillation tool (Coordinate: @user_047db929/chat-distill-pro, Version 1.0.0, Category: Knowledge Management, Publisher: user_047db929) that distills messy multi-turn WeChat/instant messaging logs, conversation screenshots, and voice-to-text content into structured 26-Scheme Prompt blueprints for AI Agents. It requires no API Key configuration. It covers scenarios such as facilitating WeChat group brainstorming, compiling scattered group tasks into requirement documents, extracting PRDs/SOPs from multi-person product discussions, and structuring voice meeting minutes. Trigger words include: Chat Log Purification, Chat Distillation, Group Chat Purification, Conversation Organization, Turning WeChat Logs into Solutions, Chat to Prompt, chat-distill, and 26-Scheme. This article introduces its positioning, core capabilities, activation methods, typical usage scenarios, and precautions.
Read Moredsh-disk-cleaner: Free Up Space on a Full Windows System Drive in DSH
dsh-disk-cleaner is a DeepSeek Harness plugin maintained by lvyue1314, designed to address insufficient system disk space on Windows. Because the DSH sandbox cannot launch subprocesses such as WizTree, robocopy, and mklink, the plugin runs in a sandbox outside the host process and provides six tools: wiztree_scan, disk_free, system_cleanup, disk_migrate, local_apps_list, and local_apps_run, with dry-run enabled by default. It is installed by linking the local package with pnpm in the profile directory and writing it into bundles or cordis.patch.yml. Requires Node >= 22 and DSH >= 0.2.1-rc.5, licensed under MIT.
Read MoreEnglish Morning Reading Courseware Generator: Transforming Word Lists into Four-Phase HTML Courseware
The "English Morning Reading Courseware Generator," published by user_08c955d3 on SkillHub (Source: community, Category: education, Version: 1.0.0, Coordinate: @user_08c955d3/english-morning-reading), is capable of bulk generating HTML courseware for 7th and 8th-grade English morning reading based on Word or plain text vocabulary lists. The courseware is organized into four stages: transcription, lead reading, sway ball, and assessment. It simultaneously outputs accompanying preview sheets and replicates the style of 8th-grade Seewo courseware, featuring three types of interactions: pronunciation reading, sway ball shadowing, and in-class quizzes. Tagged with `requires_api_key=false`, it does not require additional API Key configuration. Installation commands were not provided in the captured material; the page only lists the coordinate and official address. Triggered by natural language, trigger words include "make morning reading courseware," "generate morning reading," "English vocabulary courseware," "turn this word list into morning reading," "morning reading class," and "new morning reading." Suitable for English morning reading for the first semester of 7th and 8th grade; input must be a vocabulary list, as grades outside this range or non-vocabulary inputs are not supported. The page shows 0 installations, so actual effectiveness requires personal verification.
Read MoreGuanyuan CLI Official Suite: Bringing BI Platform Operations into the Command Line
Guandata has released the entry point for its CLI suite at @org-v7o6v577/guandata-cli-suite (version 1.0.3, categorized as data-analysis and sourced from enterprise). It consolidates ETL, datasets, pages, cards, metrics, lineage queries, form-based data entry and writing, SuperApp template creation, data source and dataset management, ETL development, chart and dashboard building, metric editing, and workflow and data pipeline orchestration on GuanBI into a single command-line interface. The operations are organized into six components—guancli, guands, guanetl, guanvis, guanmetric, and guanwf—where users select the appropriate component based on the task. The suite is tagged with requires_api_key=false, meaning that using the Skill itself does not require an additional API key. If the components are not installed, it guides users to install the official suite via @guandata/guanskill. This setup is suitable for scenarios requiring batch operations, scripting, and auditable replays. Users should note that the installation volume is still relatively low, the authentication configuration for connecting to the platform is not specified in the documentation, and cross-component tasks require users to orchestrate them manually.
Read MoreYishu Memory Lattice: DSH Memory Plugin with Lifecycle
Yi Shu (memory-lattice) is a memory plugin developed by Lumivox-wiz for DeepSeek Harness. It features 48-hour short-term expiration, long-term review based on the SM-2 algorithm, and dormancy with wake-up capabilities for unused memories; LTM writes are routed through an audit center. The system utilizes dual-channel retrieval (local BM25 and vectors), versioned personality profiling, a 3D memory nebula, encryption, incremental backups, and recycle bin recovery. Based on the official README, this article covers installation commands, typical usage, and application boundaries.
Read Moredsh-delivery-kit: Generate Delivery Acceptance Materials from Repository and Running Systems
Dsh-delivery-kit (Delivery Workshop) is a DSH workflow plugin released by Lucky020. Without uploading code, it generates eight deliverables locally from repositories and running systems, including database design, high-level/detailed design, user manuals, test reports, deployment reports, acceptance reports, and technical agreements. It executes through two steps: delivery_catalog and delivery_doc, returning "missing" for unavailable materials without fabricating content. Installation: `dsh plugin --profile web add github:Lucky020/dsh-delivery-kit`. Dependencies: python-docx and Chrome, Node >= 22, MIT License, version 0.3.5.
Read More