dsh-google-chrome-use: Enable DSH to use local Google Chrome
Maintained by BaronCyrus, the DSH networking tool plugin `dsh-google-chrome-use` (0.2.1) is built on `chrome-devtools-mcp`, allowing models to open webpages, interact with them, and take screenshots on a local Google Chrome instance. It supports both independent configuration and connections to an existing daily Chrome profile. This article covers the two operational modes, the official installation command (`dsh-google-chrome-use@0.2.1`), connection methods, configuration essentials for `cordis.patch.yml`, as well as token usage and security considerations. It is compatible with macOS and Windows.
Read MoreDe-AI-ifying Professional Texts: Rewriting and Quantitative Detection for Technical Scenarios
A Skill for De-AI-ification of professional texts (e.g., academic papers, official documents, and reports): During rewriting, it preserves terminology, data, citations, legal provisions, formulas, and academic hedging, while ensuring that normative terms (such as "shall", "should", and "shall not") are not mistakenly altered. It supports three intensity levels (light polishing, moderate rewriting, and heavy restructuring), 12 styles, mixed Chinese-English content, and batch rewriting. It includes `scripts/ai_score.py` to output a 0–100 four-tier score, a hit list, and before-and-after rewriting comparisons. For users without a Python environment, a web-based detector and prompt-based fallback solutions are available. It also features human-written sample anchors, contamination self-check, and troubleshooting and correction mechanisms.
Read Moredsh-pitfall-memory: Distilling Tool Failures into Reusable Long-term Lessons
dsh-pitfall-memory is a DSH memory plugin maintained by baizhu66018. It captures tool failures at runtime, generates lessons and rules through a four-layer convergence process, and injects briefings via systemPrompt.context in subsequent sessions. It supports session/daily/weekly summaries, pitfall_* tools, and a Web settings page. To install, run node scripts/install.mjs and restart DSH. Configuration is defined in cordis.patch.yml, with data stored in .dsh-pitfall-memory under the profile directory. This is not a replacement for AGENTS.md. Released under the MIT license; requires Node.js >= 20.
Read MoreA-Share Asset Overview: Porting the Choice "Asset Overview" Sheet to the WorkBuddy Preview Panel
Introducing the community Skill "A-Share Asset Overview" (@user_058048a4/asset-overview, v1.1.0). It replicates the "Asset Overview" sheet found in the Choice individual stock diagnosis module. By inputting a stock code or name, it fetches data from the Eastmoney balance sheet and income statement to calculate asset composition, liabilities and shareholder equity, capital sources and utilization, surplus funds, OPM, leverage, and debt service indicators. It outputs a self-contained HTML report featuring 14 inline ECharts graphs, which is displayed in the WorkBuddy preview panel. No API key is required, and trigger words include "Asset Overview", "Balance Sheet Overview", and "Balance Sheet Analysis".
Read MoreDSH-DREAM: Agent-Oriented Local-First Long-Term Memory Layer
DSH-DREAM is a DSH memory plugin and Python library maintained by baiqieji-LuLu, employing holographic writing and four-layer context separation, with default local BAAI embeddings and Chroma+BM25/RRF retrieval. The article describes the Pre-Alpha installation script, Settings options, Holographic mode, and RAGStore usage, while cautioning about plugin permissions and noting that the community directory is not an official store.
Read MorePaper Harness (Literature Station): PDF Deep Reading Workbench on DeepSeek Harness
Paper Harness is a client plugin developed by baigaobai for DeepSeek Harness. It automatically indexes PDFs added to a local library and features a three-column interface that supports selecting sentences from the original text to query AI, streaming translation by page, and management via categories, tags, and status. The library supports full-text search locally, as well as one-click importing from online searches on arXiv, Semantic Scholar, and Crossref. Assistants can read and write to the literature library via the `paper_library` tool. To install, clone the repository and run `pnpm install`, then add it to your profile via a bundle link or `plugin_manager`. Both data and translations are stored locally in the `library/` directory. Released under MIT 0.1.0.
Read MoreProject Closing Deliverables Generator: From Proposal to Finalization Trilogy
The Research Project Conclusion Package Generator is a writing and organizing Skill designed for the project conclusion phase. Published by community developer user_6ab891d8, it is categorized as "professional" with version 1.0.0. Starting from the project initiation proposal, it generates a one-stop "conclusion trio": research questionnaires (including interview outlines and implementation plans), initial drafts of concluding journal papers, and literature supplements along with reference verification reports. It applies to the conclusion stages of provincial and municipal philosophy and social science planning projects, educational science planning projects, and vertical projects such as those under the State Ethnic Affairs Commission. To install and use, run `skillhub install @user_6ab891d8/gyc3`. The tag `requires_api_key=false` indicates that no separate API Key application is needed during operation. The article introduces core features, installation and activation, typical usage, applicable scenarios, precautions, and provides the official address.
Read MoreStatistical Analysis MCP Server: Integrating Hypothesis Testing and Regression Analysis into 30 MCP Tools
## Introduction In data analysis Agent workflows, statistical testing is a common requirement: determining if there is a difference between two group means, whether categorical variables are correlated, or if regression coefficients are significant. While these calculations can be performed using Python's scipy and statsmodels, Agents often need to write scripts, piece together functions, and handle output formats on the fly. The Statistical Analysis MCP Server pre-configures these capabilities into 30 MCP tools, allowing Agents to invoke them directly by tool name without having to write statistical code for every task. ## What is this The Statistical Analysis MCP Server is a community-maintained MCP server with the coordinates @user_053d27d0/mcp-stats-engine-zx. It is published by user_053d27d0, version 1.
Read MoreProcessing Tables and Word Documents into Template Styles (Lightweight Package Available): Reorganizing Files to Be Processed According to the Template Table
Guided by a template table, this Skill aligns source .xls/.xlsx files or Word-based paragraph question banks (.docx) with the template's column names, column order, and data formats. It outputs a new worksheet named "Output" by default, leaving the original file unchanged. Key features include splitting combined options within a single cell (e.g., "A-x|B-y" or "A.Text B.Text") into separate columns (Option 1/Option 2), conditionally deleting entire rows (e.g., removing rows where Question Type = Short Answer), and dynamically expanding option columns to prevent data truncation. Detailed conversion errors are outputted separately for review. Proficient use of this Skill facilitates the restructuring of question banks, questionnaires, and import templates.
Read Moredsh-teamai-panel: Manage TeamAI in the DeepSeek Harness Sidebar
dsh-teamai-panel is a DeepSeek Harness client plugin released by backtomyfuture. It completes teamai installation and team repository initialization through the "TeamAI Resource Center" sidebar, displays three skill lists covering team/active/DSH usage, maintains mcp/mcp.yaml and prompts/ with single-file commits, and invokes local commands such as teamai pull and mcp inject. Installation requires running teamai init first; on Desktop, use dsh plugin --profile desktop add dsh-teamai-panel. Note that usage records may show 0 under workspace-write. Licensed under MIT, version 0.3.2.
Read Moredsh-resource-center: Unified Management of LiteLLM Gateways, Models, MCP, and Skills in the DSH Sidebar
dsh-resource-center is a DSH community plugin maintained by backtomyfuture. Located in the sidebar under "LiteLLM Resource Center," it uses Virtual Keys to centrally manage LiteLLM activation, selected model writing, gateway MCP installation, skill installation to ~/.dsh/skills, and model-based consumption statistics. To install, fully exit DSH and run `dsh plugin --profile desktop add dsh-resource-center`. The key is stored in the credential `DSH_LITELLM_KEY`; DSH must be restarted after configuration changes. Licensed under MIT; tested against DSH 0.2.1-rc.5.
Read MoreValue Investing Stock Screening Tool (Market Earnings Rate Quantitative Analysis + Buffett Qualitative Analysis): Initial Screening by Market Earnings Rate and Qualitative Evaluation of Economic Moat
This Skill combines Eric Ward's Price-to-Earnings Ratio (PR) valuation system with Warren Buffett's qualitative analysis framework to create a dual-dimension stock selection workflow. It first performs a quantitative preliminary screening using PR metrics, applying dividend and ROE adjustments to targets requiring modifications (such as cyclical stocks and bank stocks), and then conducts a qualitative assessment using moat analysis. Published by SkillHub user user_4c93aead, located at @user_4c93aead/jztz1-skill. Source: Community | Category: Professional | Version: 1.0.10 | Tag: requires_api_key=false (No API key required). The official description details an "evolution engine" featuring a cumulative knowledge base (industry benchmarks, red flag lists, false negative library, false positive library), a "zero-omission" four-fold guarantee, an evidence-driven rule revision protocol, and mandatory standardized profile outputs. This article is based on verified data. The official page failed to be retrieved during this capture; please refer to the official page for information regarding advanced capabilities.
Read MoreSkill Directory Generator: Batch Parse SKILL.md to Output Markdown and JSON Directories
A Skill catalog generator sourced from the SkillHub community (@user_211e8f0b/skill-catalog-generator, version 1.2.0, categorized under dev-programming, no API Key required), designed to solve the problem of catalog maintenance difficulties as the number of skills grows. It scans all skills within a specified directory, intelligently parses the frontmatter (name, description, version, etc.) of each SKILL.md file, and classifies them by function using a multi-dimensional classification algorithm based on keywords and semantics. It generates a SKILL_CATALOG.md containing a directory index, classification details, and statistical information, as well as a machine-readable SKILL_CATALOG.json. The input directory convention assumes one subdirectory per skill under .trae/skills/, containing a SKILL.md file. It is executed via `python scripts/generate_skill_catalog.py`, with options to specify the directory using `--skills-dir` and output both formats simultaneously using `--format both`. It supports automatic updates via scheduled tasks (Windows/Linux/Mac) and features comprehensive logging and error handling. This tool is triggered when users need to generate skill catalogs, classify skills, list skills, organize skills, perform batch management or archiving of skills, or auto-generate skill documentation. Please note that for skills with ambiguous boundaries, the classification results may place them into adjacent categories, so manual review is recommended. Additionally, parsing relies on the completeness of the frontmatter.
Read MoreAnalysis Report on Construction/Engineering Labor Costs: Composition of Per Capita Labor Costs and Industry Benchmarking
This article introduces the Architecture/Engineering Labor Cost Analysis Report skill on SkillHub, provided by Ant HR. Version 1.0.0, categorized under data-analysis, and no API keys are required. It breaks down the per capita labor cost in the architecture/engineering sector into five categories—wages, social insurance, housing provident fund, benefits, and administration—provides industry benchmarking and cost-reduction recommendations, accepts input via natural language prompts or by uploading files, and, when dynamic data definitions are unavailable, prompts for verification rather than fabricating results. It outlines the core capabilities, input methods, verification of installation and activation, typical use cases, applicable scenarios and limitations, and provides the official link.
Read Moredsh-temp-snapshot-cleaner: Clean up DSH orphaned workspace snapshots in the system temporary directory
DSH's dsh-workspace-changes creates dsh-workspace-changes-XXXXXX snapshots in the system temporary directory; abnormal termination leaves orphan directories that consume significant disk space. dsh-temp-snapshot-cleaner is an admin-security interface plugin maintained by b9p79rgh6f-sudo, which scans the temporary directory via the sidebar "Cache" entry and deletes after five safety checks, a dry-run manifest, and secondary confirmation. The default retention period is configurable at 72 hours, with manifests and logs stored in ~/.dsh/plugin-data and ~/.dsh/logs. Installation is available via dsh plugin add dsh-temp-snapshot-cleaner or GitHub source, MIT 0.1.0.
Read MoreWeChat to ima Assistant: Automatically Save Articles to Knowledge Base
The WeChat Article to ima Assistant (@user_00bc0e74/wechat-article-to-ima, v1.0.0, a community-sourced AI Agent) identifies mp.weixin.qq.com/s links, article cards, or content combinations containing trigger phrases like "转存ima," "存入ima," or "存到ima" within conversations. It saves these articles to the "琪儒的知识库" personal knowledge base in ima. Supporting both single and batch saving (up to 10 links at once), it features cross-user compatibility but does not support automatic tagging. Credentials are configured locally by the user in ~/.config/ima/; the Skill does not contain built-in credentials or Knowledge Base IDs. The target knowledge base is located dynamically by name, ensuring no credentials are leaked when shared. The article covers trigger scenarios, core capabilities, privacy and credential mechanisms, installation and activation, typical usage, applicable scenarios, limitations, and provides the official link.
Read Moredsh-balance-widget: DSH Sidebar Balance and Usage Widget
This article introduces the community client plugin dsh-balance-widget (v0.3.2, MIT), which displays DeepSeek account balance, today's token usage, and estimated spending at the bottom of the DSH Web sidebar. The panel provides conversation-level breakdown, peak/idle pricing tags, and 7-day daily average progress. Data is sourced from the deepseekAccount and conversation usage event cache. Installation is performed via dsh plugin add @b9p79rgh6f-sudo/dsh-balance-widget or via GitHub source, and requires DSH Web 0.2.x+ and a logged-in session.
Read MoreCron Mastery: Heartbeat drifts; schedule tasks with cron
Cron Mastery is an OpenClaw scheduling system usage guide (v1.0.3, MIT-0, compatible with OpenClaw 2026.2.15+) published by i-mw on ClawHub. It delineates the boundary between heartbeat and cron, provides cron:add payload templates for reliable one-time reminders, distinguishes between agentTurn push and systemEvent silent mode, and covers Janitor periodic maintenance, timezone locking, Self-Wake rules for cross-round waiting, legacy task migration, and common troubleshooting.
Read Morehuahua-purge: Defense Evasion Engine and Red Teaming Platform on DSH Hosts
huahua-purge is a bundle host plugin for DeepSeek Harness, maintained by azure5100 and categorized under admin-security. It combines armor breaching (stripping host identity and persona injections, switching prompts by model, anti-bypass hooks, and automatic retries) with a red team exercise platform (five-role orchestration, 23 skills, SQLite asset library, and HTTP bridge) into a single installation. It requires DSH ≥ 0.2.1-rc.2 and Node ≥ 20. The installation command is `dsh plugin --profile desktop add "github:azure5100/huahua-purge"`, and the host must be restarted after installation. It is intended for authorized penetration testing and internal drills; please review the source code and the MIT license before use.
Read Moredsh-smooth: Using a configuration bundle to alleviate DSH lag, reconnects, and slow startup
dsh-smooth (ayuanxx-max, v1.2.0, MIT) is a pure configuration bundle plugin for the DSH client. It adjusts UI compact mode, jobs polling, and other settings via cordis.patch, and uses optional documentation to explain that on desktop, compression thresholds need to cover agent presets, tool slimming, Agents-Anywhere reconnection, and sub-agent removal. Installation can be done using `dsh plugin --profile desktop add dsh-smooth` or through profile package.json linking; changes take effect after completely exiting DSH, and removing it from bundles allows for rollback. Maintainer test data is available for Windows 11, while Linux/macOS remain unverified.
Read MorePC Single-Player Game Review System: From Game Package to Structured Evaluation Report
This is a community Skill on SkillHub published by user_354646e6. It is Version 2.0.0, categorized under content-creation, utilizes the universal SKILL.md format, and requires no API Key. Targeting PC single-player games, it automatically identifies 12 game genres and applies a specialized analysis framework for each to deconstruct core mechanics, analyze system coupling, and perform multi-dimensional scoring, ultimately outputting a structured Markdown evaluation report. Supported genres include Strategy, Action, RPG, Shooter, Platformer, Simulation, Survival, Puzzle, Adventure, Rhythm, Racing, and Horror. Official documentation does not provide installation commands or complete invocation examples, offering only trigger words (游戏评测, 游戏分析, 玩法拆解, game review, game analysis). It is suitable for developers requiring the batch generation of structured single-player game reviews. Note that scores are not directly comparable across different genres, and the evaluation of numerical balance relies on the quality of the package data.
Read MoreNi Haixia's Constitutional Dialectics: Prescription Study and Wisdom Nourishment – University Edition: A Dual-Track TCM Consultant Skill for University Scenarios
The Skill "Ni Haixia · Body Identification and Syndrome Differentiation for Prescription Wisdom and Health Cultivation - University Edition" (User ID: @user_a84e2069/heda-nishi, v1.6.0, Category: community/life-service, requires_api_key=false) is a dual-track Traditional Chinese Medicine (TCM) advisor designed for any university. It combines "Health Cultivation" (covering the Eight Principles, Zang-Fu organs, Qi-Blood-Body Fluids, and Six Dimensions of Health adapted for student schedules, sedentary study, cafeteria diets, and dormitory environments) with "Classical Prescriptions" (based on Ni Haixia’s *Ren Ji*, Six Meridian Syndrome Differentiation, and classical formula references). Additionally, it features deterministic scoring for nine body constitutions based on GB/T 30318-2013 and generates HTML visual reports. Anchored in 16 ancient medical texts, including the *Yellow Emperor's Inner Canon*, *Treatise on Cold Damage*, and *Synopsis of the Golden Casket*, the system integrates 14 in-depth study modules, 1,257 clinical cases, dosage tables for 99 formulas, and an acupoint quick-reference guide. It requires an exact match of the university name to confirm the campus location and adjusts recommendations regarding the four seasons, solar terms, dietary therapy, and moxibustion intensity based on the seven major climate zones of the city. As of November 4, 2026, SkillHub statistics show a rating of 100,000, 1,057 downloads, and 0 installations. This article outlines the core capabilities, trigger words, and typical use cases, noting that the documentation lacks official installation commands and the skill requires no API Key.
Read MoreHydraulic Pump CAD Drawings: A Skill for Delegating Drafting to an Agent
The "Hydraulic Pump CAD Drawings" is a community-sourced design-media Skill located at `@user_5453610a/hydraulic-pump`. Version 1.0.1, released by `user_5453610a`, has a page rating of 100,000 and 1,058 downloads (0 installations). It is tagged with `requires_api_key=false`, meaning no additional keys are required. Its core capability is the generation of hydraulic pump CAD drawings, triggered by intent matching. The official specifications include four trigger types: the Chinese short phrase "液压泵", the command "生成液压泵图纸", the colloquial "做一个液压泵", and the English identifier "hydraulic-pump". Covering Chinese, English, colloquial, and written language, it allows users to operate without memorizing fixed commands. Its value lies in streamlining the drawing initiation process: traditional workflows require manual selection followed by opening CAD software to draw the pump body interfaces and mounting surfaces, necessitating redraws whenever parameters change. This Skill consolidates the entry point into a single natural language statement, enabling the Agent to generate drawings directly within the conversation. It is suitable for preliminary design drafts, iterative parameter adjustments, and scenarios requiring drawing files for secondary processing. However, boundaries must be clarified: it is not a CAD kernel plugin and does not perform strength verification, interface matching, or simulation calculations; the rationality of the selection must still be confirmed through existing workflows. Ambiguous trigger words may result in multi-turn follow-up questions. Since the installation count is 0, community feedback cannot yet be verified from existing data, so it is recommended to conduct a small-scale trial before integrating it into formal drawing workflows. The official documentation does not provide installation commands or explain the differences between Cursor, Codex CLI, and Claude Code; this article states facts without speculation and provides the SkillHub page and API address for verifying the installation method of the current version.
Read MoreResearch & Analysis Assistant: Turning Fragmented Questions into Source-Citation Research Deliverables
The "Research & Analysis Assistant" (Ready-to-Use Research & Analysis) is a third-party Skill published on the SkillHub community at the coordinate @user_af163eaa/z-research-analysis, version 1.0.1, categorized under data-analysis. It addresses the pain points in research tasks where there is no fixed structure and outputs are difficult to verify: transforming a simple "I want to understand XX" into structured, traceable, and conclusive research deliverables. It supports five types of research: market, competitor, user, industry, and policy. The output follows a fixed sequence—conclusions first, followed by supporting arguments, with each conclusion annotated with its source. No API Key is required. This article explains its positioning, verified core capabilities, and the activation method based on the SKILL.md format. It also warns that: the installation command and configuration fields could not be verified on the public page, so the official page should be consulted; source annotations depend on search results, so important conclusions still need to be verified by opening the original links; and outputs for highly regulated domains should only be used as leads.
Read Moredsh-free-search-lite: Lightweight Web Search and Content Extraction without API Keys
dsh-free-search-lite is a community-maintained web search tool plugin for DeepSeek Harness. It provides two native tools, `free_search` and `free_search_fetch`, requiring no API key, Python, browser, or MCP subprocess. It performs parallel searches via 360 and Bing RSS with a DuckDuckGo fallback, and supports character-based pagination for web page content. It has been verified with dsh 0.2.1-rc.3 and Node ≥24.14.0. Desktop installation is available via a GitHub URL or the v0.1.1 tgz package; the Web profile can be installed using `dsh plugin --profile web add`. Please review the network egress, caching, and security limitations before use.
Read More