DX-S3-04 Interdisciplinary Consultation: A Cross-Departmental Consultation Skill for Doctors and Pharmaceutical Companies
DX-S3-04 Interdisciplinary Consultation is a professional service Skill published by Baitong AI Service Platform on SkillHub, with coordinates @baitongai/dx-s3-04-kua-xue-ke-hui-zhen, version 2.2.5, source enterprise, category professional, and tag requires_api_key=false. It targets the specific need of "interdisciplinary consultation" for doctors or pharmaceutical companies, officially marked with a coordinate system indicating doctor supply-side blind spot S3 cross-department promotion and cross matrix P4L2. The page records show 793 downloads, 0 installations, and a rating of 100000. Due to the dynamic rendering of the official page which prevents content scraping, this article only includes verified information, makes no speculations about installation commands, directory conventions, configuration fields or invocation examples, and outlines considerations regarding enterprise source, limited documentation, and medical compliance in the applicable scenarios.
Read Moredsh-notion: Using Six Tools in DSH to Manage Notion
dsh-notion is a community plugin that provides DeepSeek Harness with six tools—notion_search, notion_get, notion_page_create, notion_page_edit, notion_page_trash, and notion_query. It is built on Notion's official ntn CLI rather than the official MCP, aiming to narrow the tool surface and reduce misselection and context overhead through find-and-replace document editing and structured JSON returns. Installation requires installing ntn first, then loading the plugin into DSH via `plugin_manager install_bundle` or a Git URL (the plugin id is `notion-tools`). Authentication is read from the token via `NOTION_API_TOKEN`, `notion-token`, or `ntn login`. This article summarizes installation, configuration, and typical parameter examples based on the GitHub README and package.json version 1.1.1.
Read MoreCICD AIops: CI/CD Operations Skill for Self-hosted GitLab / Gitea
CICD AIops is a community-sourced operations skill (Coordinate: @user_993b98b1/cicd-aiops, version 0.14.0, category: it-ops-security, source: community) targeting self-hosted GitLab or self-hosted Gitea CI/CD servers. It provides a unified view of instance versions, project storage, pipeline and job traces, the runner fleet, merge requests and branch protection, and artifact inventories. It features four flagship RCA (Root Cause Analysis) scenarios: pipeline failures, runner health and queues, artifact/storage bloat, and stale work. Write operations support retrying/canceling pipelines, pausing/resuming runners, deleting artifacts, and updating branch protection; all writes are constrained by a built-in governance harness (auditing, policies, token budgets, rollback, and risk grading). An API key is required for usage (requires_api_key=true). It applies exclusively to self-hosted instances; GitLab.com / Gitea Cloud SaaS and Kubernetes deployment status are not supported, for which k8s-aiops should be used.
Read Moredsh-workflow-panel: DSH Read-Only Workflow Panel
dsh-workflow-panel is a DSH client plugin maintained by tucaogege (v0.1.1, MIT license). It provides a read-only workflow panel in the bottom-right corner of the desktop version, displaying an overview, G0–G5 gates, milestones, organization templates, task flows, and personnel cards. Data is sourced from project.json, with only GET requests issued and no file writes performed. To install, declare github:tucaogege/dsh-workflow-panel in the profile's package.json and add it to the bundles. Node.js ≥ 20 is required. The project file can be optionally specified via projectPath or the DWP_PROJECT_PATH environment variable.
Read More@ttxl314/dsh-ds-peak-valley: Display Text Peak/Valley Period Badges in the DSH Conversation Bottom Bar
The community DSH client plugin @ttxl314/dsh-ds-peak-valley (v0.1.1, MIT) displays Wen Peak/Wen Valley badges next to the statistics line in `conversation.composer.dock`. Calculating locally based on the Asia/Shanghai timezone and the DeepSeek pricing footnote windows, it refreshes every minute without requiring any API or configuration. It supports web/desktop profiles and can be enabled via `dsh plugin link` or by editing the profile manifest. It is a pure client-side solution with no telemetry.
Read MoreOmni-Platform Copywriting Adapter: One Rewrite for Xiaohongshu, Official Accounts, Zhihu, and Weibo
# Introduction The same product description requires short sentences with emojis for Xiaohongshu, logical chains for Zhihu, and full paragraphs for WeChat Official Accounts. Manually rewriting content for each platform is time-consuming and prone to losing key information during the adaptation process. The "Omni-platform Copywriting Adapter" introduced below saves this manual effort: it takes a single original text as input and outputs platform-tailored versions while keeping the core message intact. # What is this? The Omni-platform Copywriting Adapter is a content creation-oriented Skill, identified as @user_4f1237ac/multi-platform-copywriter. Published by community developer user_4f1237ac, it is currently at version 1.0.0, categorized under content-creation, and sourced from the community.
Read MoreProperty Management Information: Skill for Dynamic Integration with MCP product-catalogue Connector
Property management information (@user_dbb977dc/mcwy-res-mgmt-001, version 1.0.1) is a general-purpose Skill released by community developer user_dbb977dc, classified under ai-agent and requires no additional API key. It addresses two pain points: client code hardcoding MCP server method names, which causes failures whenever the server changes them, and connector disconnections that require manual retries. The approach is to first connect to the MCP product-catalogue; if the connection fails, it automatically requests WorkBuddy to reconnect, up to 6 times—intervals are 10 seconds for the first 2 attempts and 30 seconds for the next 4; once connected, it uses ToolSearch at runtime to dynamically query the available tool list, then matches user intent based on tool descriptions and parameter signatures without hardcoding any specific method names anywhere in the process, thereby adapting to server-side methods being added, removed, or changed. It is triggered when users ask questions related to the product catalog, product queries, or property repairs. Public materials do not provide a clear installation command or configuration fields; the official page and API endpoint could not be accessed or verified at the time of writing, so the text only lists verifiable coordinates, version, source, and the requires_api_key=false tag, and does not fabricate installation steps. Key notes include: public data shows 0 installs, making it an early-stage community Skill, and it is recommended to verify it in a test environment first; the reconnection logic depends on WorkBuddy capabilities; and intent-matching quality is affected by the quality of server-side tool descriptions.
Read MoreHeadhunter Recommendation Report Template Customization: Auto-Populate Variable Fields into a docx 免费行 Bar
Headhunter recommendation reports have a fixed format, but their fields must be manually replaced every time, and editing copied drafts often leads to omissions. This article introduces the "Customizable Headhunter Recommendation Report Template" Skill from the SkillHub community (@user_b06ad2b7/headhunter-recommend-report-skill, v1.0.0, classified under "office-efficiency", requires_api_key=false): it automatically parses the docx template of the recommendation report, identifies the variable fields within it, and then generates a new docx recommendation report based on candidate data with a single click. The article explains its three verified capabilities and applicable scenarios, and points out limitations such as the page not providing installation commands, field formats, and examples.
Read MoreWorkbuddy Daily Points: Calling the Official Check-in API Using Local Login State
**Input Information:** This information is for scenarios where users forget to manually check-in and their consecutive attendance days are interrupted. It introduces the "Workbuddy Daily Points Skill," which leverages the local login state to directly call the official WorkBuddy check-in API to claim daily points. It is idempotent, operates without UI interaction, and requires no scanning. Please verify its trigger scripts and capabilities against the introduction on the official address page, and explain limitations such as applicable scenarios and login state dependencies. For installation and activation, provide only the official address. Do not speculate on differences regarding Cursor, Codex CLI, or Claude Code, as the main body of the page could not be retrieved.
Read Moredsh-quote-selection: Box-select Text in DSH Conversations with Source Reference
dsh-quote-selection is a DSH client plugin maintained by Tsutsutsuu: it lets you select text in the chat view; using the right-click menu or Ctrl+Q inserts the selection as a chip into the input box, and once sent, the model and conversation log receive the full source and original text formatted as 【引用 · …】 (quotation · …). It covers installation (plugin_manager install_bundle), settings, and the wire text form, as well as opportunistic compatibility with the dsh-better-display reading view and known limitations.
Read Moredsh-mihealthlink: Integrating Xiaomi Mi Fitness into DSH Local Memory
dsh-mihealthlink is a DSH memory plugin maintained by Tsurukyuha. Based on mi_fitness_data_bridge, it writes data such as steps, sleep, and heart rate from the Xiaomi Cloud (China region) into a local SQLite database, and provides Cordis tools including mi_health_status, sync, report, and brief, along with a tools command line. Installation requires Python 3.11+, WSL2 or Linux/macOS, a link mount in the profile, and config.local.json, with credentials imported via keyring. Data is not uploaded. It uses non-official interfaces, so you should assess the account and privacy risks yourself before use.
Read MoreA Design Guide for Browser Automation Agents: Ensuring Reliable and Controllable Web Operations
The "Browser Automation Agent Design Guide", released by Kunlun Growth as Version 1.0.0 and categorized under "ai-agent", is built upon the universal SKILL.md format. Rather than being an executable script, it serves as a comprehensive design methodology. The guide offers in-depth coverage of several critical areas, including decision-making between AI-driven and script-driven approaches, page perception and action planning patterns, handling anti-crawling mechanisms and login states, as well as stability engineering and safety guardrails. The primary goal is to empower AI to autonomously perform web operations such as form filling, data scraping, routine inspections, and price comparisons, ultimately advancing solutions from the demo phase to a reliable and controllable state. Notably, the guide does not require a separate API key application. This article breaks down its four major thematic modules and outlines the applicable developers, important considerations, and official channels for access.
Read MoreDomain Authority Auditor: Evaluating Domain Credibility with 40 Checks
The Domain Authority Auditor (v13.0.0), maintained by community author user_2c0f1e77, is categorized under 'data-analysis', sourced from the community, and tagged 'requires_api_key=false'. It evaluates domain authority, website trust, and citation source credibility by applying 40 CITE scores in conjunction with veto checks, concluding with one of three tiers: TRUSTED, CAUTIOUS, or UNTRUSTED. The official overview explicitly states two scope limitations: it does not perform page-level content quality assessment (for this, use the 'content-quality-auditor') and does not provide standalone offsite link profiling (for this, use the 'offsite-signal-analyzer'). This article presents its positioning, core capabilities, installation entry point, and trigger scenarios. It also acknowledges the absence of installation commands, CITE sub-score details, and reproducible examples in the retrieved materials, thus avoiding speculation that could be mistaken for fact.
Read MoreDSH Stylebook: Customizing the Font and Tri-color Theme for DeepSeek Harness
dsh-stylebook is a DeepSeek Harness appearance plugin maintained by tsuixl (version 0.1.0-beta.1, compatible with DSH 0.2.1-rc.5). It supports separate configuration of interface and code fonts, three-color primary derived themes, and light, dark, and naming schemes. Install the .tgz via the desktop CLI and use it in "Settings → Appearance". Windows desktop validation is the standard, and disabling the plugin allows reverting style overrides.
Read MoreParents' Homework Correction Workbench: A Directory of AI Grading, Error Archiving, and Daily Reports
The "Homework Grading Workbench for Parents" is a community "Skill" located at @user_5e6068f7/zuoyepigai, version 1.0.0, category "education", and tagged with requires_api_key=false. It solidifies the workflow of "photo capture and submission -> AI-based grading -> automatic error archiving -> daily practice repetition -> daily report generation" into a reusable product: Subject specific grading rules, CSV files for Mistake Books (错题本), Mistake Practice & Dictation Check Mobile Apps, and Automated Today's Report Configuration & Directory Structure. This article introduces its Positioning, Core Capabilities, Authentication Credentials for Activation (Coordinates, Version, Source, Tags, Page URL), and provides four typical request areas: Subject Grading, Mistake Archiving, Practice Check-in, and Daily Reports. The materials do not provide installation commands and platform-specific differences for Cursor/Codex CLI/Claude Code. The article does not infer these differences, but states that the SkillHub page should serve as the definitive reference. It reiterates that this is a "Skill" for building the work platform, the grading criteria still require parental review and verification, and subject-specific rules must be generated separately for each discipline.
Read MoreR&D Expense Compliance Diagnosis: Pre-filing Compliance Check for R&D Super Deduction
# Introduction Before applying for the additional deduction of R&D expenses, financial personnel typically review the enterprise's qualifications, the authenticity of R&D activities, expense accumulation, commissioned R&D, and documentation for inspection item by item. Relying on manual ledger checking and spreadsheets often leads to high-risk points being overlooked, with issues only discovered during the final tax settlement or upon audit. The Skill introduced below transforms this type of compliance check into a structured Q&A process. By inputting the enterprise's situation and project data, it outputs a risk level, a list of issues, and potential tax consequences. # What is this Skill Skill Name: R&D Expense Compliance Diagnosis, located at `@user_724120d6/axy-dev-dx`, Version 1.0.4. Published by `user_724120d6`, Source: community, Category: professional, Tags: `requires_a`
Read MoreMiao Shou ERP-AI-powered Product Polishing: An AI Polishing Feature for Cross-border E-commerce Product Titles, Descriptions, and SKU Specifications
Miaoshou ERP-AI Product Polish is a product copywriting Skill designed for cross-border e-commerce, utilizing Miaoshou ERP's public AI services to generate or refine product titles, descriptions, and SKU specifications. Scraped data indicates its identifier is @org-ko8itrop/miaoshouerp-ai-product-polish, with version 1.0.1. Sourced from enterprise, it is published by Shenzhen Chengyun and categorized as business-ops, with the tag requires_api_key=true. Capabilities include rewriting or translating titles/descriptions, keyword inclusion, negative word exclusion, image-assisted copywriting, AI/language queries, and SKU specification standardization, covering platforms such as TikTok, Ozon, Temu (fully managed), Temu (semi-managed), Shopee, and MercadoLibre. As specific installation commands and configuration fields are not provided, users must prepare an API Key beforehand and refer to the official page for accurate instructions.
Read MoreDeepSeek Time-Specific Price Widget: Peak/Off-Peak Rates, Balance, and Project Spending within Harness
Developed and maintained by TSRYWHY, the DSH client plugin **deepseek-time-price** provides a draggable floating window and an input box entry within the Harness Web interface. It displays DeepSeek's peak/idle time slots, a 24-hour progress bar, Platform balance, and cost estimation for projects in the same cwd. Additionally, it fetches China's statutory holidays via a Host read-only route to assist in time slot determination. Installation is performed via the Plugin Manager's `install_bundle` and `cordis.patch.yml`; it is applicable to local loopback Harness and is released under the MIT license.
Read MoreNew College Entrance Exam Class Scheduling Skill: Handling 3+1+2 / 3+3 Elective Course Scheduling via Constraint Solving
"New Gaokao Course Scheduling" is a community skill package targeting the universal SKILL.md format, with the coordinate @user_a16b2e76/xingaokaopaike, version 1.0.0, and categorized under education. It models the "administrative classes (Chinese, Math, and English) + mobile classes (elective subjects)" system under the New Gaokao (3+1+2 / 3+3) framework as a constraint satisfaction problem, automatically generating conflict-free administrative class schedules, teaching class schedules, and teacher schedules. The scheduling objectives are student conflict-free, teacher conflict-free, and classroom alignment. It is designed for secondary school academic affairs staff implementing the mobile class system, or for developers embedding course scheduling into Agent workflows. The article introduces its capability boundaries, installation entry points, typical workflows, and precautions for use, and reminds users that third-party community skills should be fully tested in a pilot grade before integration.
Read Moredsh-monokai-pro: Applying the Monokai Pro Theme to DeepSeek Harness Web GUI
Maintained by trespassmk, dsh-monokai-pro injects the official Monokai Pro color palette (derived from VS Code Extension 2.0.16) into the DeepSeek Harness Web GUI as a client bundle. It offers six dark filters (including Pro and Classic), syntax highlighting, full chrome support, and a Monokai Pro Light theme. Install via `dsh plugin --profile web add dsh-monokai-pro`; toggle and select colors under Settings → General in the Monokai Pro row. Settings and filters are stored in localStorage. MIT License, version 1.1.1.
Read MoreDeeply De-AI Your Copy: Turning Robot-Sounding Text Back into Natural Human Voice
Introducing the community Skill "Aggressive AI Text De-flavoring" (@user_43d6a87b/quaiwei, v1.0.0, content-creation). Addressing the problem of Chinese and English manuscripts that "read like AI," the workflow first utilizes detection scripts to statistically assess clichés, mechanical "first/second/last" structures, dash usage, and syntactic regularity, generating 3–5 diagnostic notes. Following this, without altering the original meaning (where numbers, conclusions, and argument directions are strictly fixed), the tool removes filler words, adds details, and rewrites by varying sentence lengths. Tone options include Casual, Professional, Sharp, and Gentle. The intensity parameter controls the rewriting strength (light mode only swaps vocabulary, while heavy mode restructures the skeleton). No API Key is required. Official documentation lacks specific installation commands and installation differences for Cursor/Codex CLI/Claude Code; please refer to the SkillHub page for accurate information.
Read Moredsh-whale-game: Launch and Stop a Local AI Playground in DSH with One Click, Playing at the Same Table as Agents
dsh-whale-game (Big Fat Fish Game Hall) is a DeepSeek Harness plugin maintained by Treasure-hub-agent: start/stop the built-in AI playground from the sidebar with one click (default 127.0.0.1:8787), with six chess/dice mini-games and a web client included; human players open a table via browser, Agents join via MCP, and rules are adjudicated by the pure-function kernel of ai-playground. The plugin provides MCP copy, an optional wakeup bridge, and MCP self-registration; the game hall runs as a detached process. Installation: `dsh plugin --profile desktop add dsh-whale-game`, requiring DSH >= 0.2.0-rc.1 and Node >= 22, under the MIT license.
Read MoreEmpower Your Prompts: Clarify Intentions Before Rewriting
The SkillHub community skill "Prompt: Give Me a Helping Hand" is designed to optimize existing prompts. It first clarifies the user's true intent through targeted questioning, then provides a single rewritten, optimized version accompanied by an explanation of the changes. It covers Agent/system prompts as well as task-oriented prompts for writing, analysis, role-playing, and code generation, without being tied to specific platforms. Trigger words include "optimize prompt," "improve prompt," "polish prompt," "help me revise this prompt," and "optimize agent prompt." It does not handle prompt creation from scratch, prompt evaluation/scoring, or text-to-image/video generation prompts. Before use, the existing prompt and goals must be confirmed. The workflow follows the path of: clarifying intent, outputting the rewrite, and explaining the changes.
Read MoreMoments Interaction Assistant: Managing Likes and Comments on WeChat via Cloud Phone
Introducing a community-sourced Agent Skill: "Moments Interaction Assistant" (@user_c0f5ae43/moments-interaction, 1.0.0). It performs Moments likes, comments, and scheduled checks on WeChat on a cloud phone. Comments uniformly follow a positive-energy praise logic, only saying positive and warm words, with each post commented on only once. Category: business-ops; requires_api_key=false. At the time of retrieval, the rating was 100000, downloads 851, and installs 0. The official documentation does not provide installation commands or usage examples; this article reports this as-is and recommends enabling it according to the Skill directory rules in each Agent environment.
Read Moredsh-yet-another-slider-thingy: Inference intensity slider on Composer model slots
The Traveritas DSH plugin dsh-yet-another-slider-thingy overrides the Composer slot conversation.input.model with priority -1, adding a discrete reasoning-effort slider while preserving the native model directory and MenuSurface. It offers three appearances (compact/block/pill), thumb shape and size options, step-wise or smooth dragging, five track effects, and editable colors, with settings centralized under Thinking Intensity. Installation: git clone the repo, then link a local absolute path via the plugin manager; verified on DSH 0.2.1-rc.5. Uninstalling or disabling the toggle restores the official model slot.
Read More