dsh-session-porter: Export DSH sessions to a sync directory and import them back on another machine
dsh-session-porter is a DSH memory plugin maintained by LZ-jun that writes sessions to a user-chosen directory (such as OneDrive) via sidebar Export/Import buttons, with support for per-session selection, an overwrite toggle, and confirmation dialogs. Exports include jsonl logs, meta, and attachments; an optional plugins.json is provided only as a reference manifest. To install, search for dsh-session-porter in dshmarket and restart DSH; configuration is specified in session-porter.json. All operations use a local loopback with no external requests. Import does not modify workspace files. Node.js ≥ 22 is required.
Read MorePerformance Self-Evaluation Writing Assistant: Transforming Work Log Drafts into STAR-based Self-Evaluations
The Performance Self-Evaluation Assistant is a writing tool designed for employees to draft their performance reviews. Published by SkillHub community user_fa14b5bc, it falls under the "office-efficiency" category, is version 1.0.0, and requires no API Key. It transforms chronological work logs into STAR-structured self-evaluations, outputting five sections: STAR rewriting (including S/T/A/R four-dimensional quality scoring), value extraction, evidence gap detection (with supplementary guidance), a complete self-evaluation draft (available in three styles: professional & powerful, modest & pragmatic, or highlight-focused), and a leader's perspective preview (highlights/concerns/blind spots). It features built-in exception handling for four scenarios (too brief drafts, lack of results, soft contribution templates, rewriting failures) and corrections for six common self-evaluation errors, covering the specific focuses of five job categories: development, product, operations, management, and functional roles. The workflow involves preparing a chronological draft, submitting it to the Skill for rewriting, identifying weak sections via four-dimensional scoring, supplementing data based on evidence detection, selecting a style to generate the final draft, and reviewing the leader's preview to ensure concerns are addressed. This Skill is responsible for text structure optimization; it does not fabricate real data. The leader's preview serves only as an auxiliary judgment tool and cannot replace actual communication.
Read MoreIntegrating COROS in DeepSeek Harness: Login and Read Sports & Health Data
**dsh-coros-sport** (lyy-hast/dsh-coros-plugin, v1.0.4) is a Cordis plugin for the DeepSeek Harness desktop client. It enables **Coros OAuth** login directly within the panel and retrieves account, device, recovery, fitness, and training schedule data—as well as data from the last 7 days of activity and health—via the official **MCP**. The **/coros-plan** command generates text-based training tables based on account summaries. The sole write operation creates a built-in 4-week, 16-session plan within the "Training Schedule → Details" section, following a preview and two-step confirmation process. The plugin supports installation via `npm install_bundle` and configuration for `region` and `overviewDays`. It connects directly to Coros clusters for data and stores tokens locally.
Read MoreVideo Copy Extraction Master: Converting Video Speech to Usable Text
The challenge of video-to-text transcription lies not in recognition, but in cleansing: removing filler words, slips of the tongue, and repetitions, then organizing the content into scripts based on intended use. This Skill released by the community (@user_fb71c45d/shipin-wenan-tiqu-dashi, version 1.0.0, category: content-creation) integrates cleansing into the model's processing workflow. It supports both network video links and local video file inputs, leveraging the Qwen LLM for semantic parsing to automatically eliminate filler words, slips of the tongue, and repetitions, while supporting the generation of copy tailored to specific styles. It is suitable for short video derivative scripts, corporate meeting minutes, online course breakdowns, live stream reviews, interview transcriptions, knowledge material extraction, store visit scripts, and training courseware organization. Please note that the tag `requires_api_key=true` indicates a valid Qwen API Key is required for activation. With a current rating of 100,000, 905 downloads, and 0 installations, this relatively new version suggests validating with real materials before deployment.
Read Moredsh-agent-team-workflow: Orchestrating Recoverable Agent Team Workflows with Node Graphs in DSH
dsh-agent-team-workflow is a DSH workflow plugin maintained by lyingloong. It defines nodes such as agent, command, HTTP, human, fan-out/join, and subworkflow in JSON using apiVersion 2. At runtime, it handles scheduling, permission grants, and atomic in-session JSON persistence, and supports lane/model failover, gates, and a web dashboard. Installation requires Node.js >= 22.5. Install with: dsh plugin --profile desktop add github:lyingloong/dsh-agent-team-workflow, then restart Harness. Typical usage is to describe the process in natural language, then call team_workflow_define/start/run/decide/status. command/http actions require user authorization. The v2 runtime has no v1 compatibility.
Read MoreLearn Constitution Differentiation, Syndrome Differentiation, and Classical Formulas with Master Ni: Dual-Track TCM Consultant Skill
Introducing the Skill “Identify Constitution and Syndrome with classical formulas, health cultivation alongside Teacher Ni” at @user_a84e2069/tcm-nishi, version 1.5.5, source: community, category: professional. It is positioned as a dual-track Traditional Chinese Medicine advisor: the wellness track focuses on the eight principles, zang-fu organs, qi–blood–fluids, and the six dimensions of health cultivation; the classical formula track follows Teacher Ni Haixia’s Human Chronicles Six-Meridian pattern differentiation and generates flashcards from classical formula clauses. Constitution assessment uses deterministic scoring based on the nine body constitutions in GB/T 30318-2013 and outputs an HTML report. The academic foundation draws on 16 ancient TCM classics and includes 14 in-depth study modules of these classics plus a clinical case repository of 1,257 cases. The provided materials do not include installation commands or configuration fields; only verified metadata and trigger scenarios are listed.
Read Moredsh-ccswitch-usage: Write DSH token usage into the CC Switch usage panel
Dsh-ccswitch-usage is a DSH plugin maintained by LyaQanYi. It monitors the token field of the assistant/message within sessions, writes data to CC Switch's proxy_request_logs (app_type=dsh), and aligns fields with the official DSH importer. It supports backfilling upon startup and cost calculation based on model_pricing. For desktop versions, installation must be performed using the app's built-in CLI. It is ideal for viewing the usage of DSH and CLIs such as Claude Code centrally within CC Switch.
Read MoreInteractive Prototyping Skill: Customizable HTML Prototypes Driven by DESIGN.md
Introduce the SkillHub community skill “Interactive Prototype (Auto-generates key annotations, customizable style)”, located at @user_9c6b6ed5/product-prototype, version 1.0.1, category design-media. It converts PRDs or requirement descriptions into interactive HTML prototypes and automatically generates key annotations based on the PRD. The core mechanism separates the structural skeleton from the visual specification: colors, fonts, corner radius, and component styles are centralized in DESIGN.md, allowing style changes by modifying the spec while keeping page structure and connections stable. It can also extract design style from a user-provided webpage link to generate or update DESIGN.md. The article presents the directory structure and configuration patterns under the SKILL.md format, along with DESIGN.md variable examples, and outlines three typical use cases: generating a prototype from a PRD, extracting style from a reference site to generate a prototype, and regenerating after a client requests a style change by altering only the spec. It also specifies the target users and notes that visual changes must go through DESIGN.md, structural rebuilding is out of scope, and installation paths should be verified in a real editor due to environment differences.
Read MoreAgent Development Assistant (Backend Development): Breaking Requirements Down Into Verifiable Development Pipelines
An 8-stage universal development pipeline Skill for backend developers, ID @user_f3f6053c/zhinengtikaifa-001, Version 1.0.1, community-sourced, Category: dev-programming, No API Key required. It crystallizes the requirements-to-commit workflow into: converting PRD + API specifications into a structured specification document, atomizing into single-file / single-function tasks, precisely locating and implementing code, enforcing two validation gates (static analysis plus configurable runtime verification), and accumulating knowledge with semantic commits. It supports three ready-made templates (Python, TypeScript, and general) that auto-adapt to the project type. The article notes that the approach is based on a workflow already proven inside Tencent and is currently community-curated. It also states that the premise is requirements and prototypes already in place, and runtime validation must be self-configured.
Read Moredsh-memos-panel: Chinese Settings Panel for MemOS Cloud Memory
dsh-memos-panel (v1.2.9) is a community-developed third-party plugin for the DSH client, providing a Chinese interface for @memtensor/memos-cloud-dsh-plugin 0.1.2. It features 47 parameters across 8 categories, dual-path connection testing, optional secondary filtering for OpenAI-compatible recall, and six local privacy rules. To install, users must first enable the official MemOS plugin and then add the package via the direct .tgz link from the GitHub Release. The configuration entry point is located at Settings → MemOS Memory. This plugin has no official affiliation with DeepSeek or MemTensor. It is compatible with DSH >=0.2.1-rc.2; users are advised to review the source code and license independently.
Read Moredsh-chat-shell: Seamless Chat and Harness Switching on the DeepSeek Harness Desktop
dsh-chat-shell is a DSH client plugin maintained by lxg103, targeting DeepSeek Harness desktop 0.2.x on Windows. It provides a Chat/Harness segmented toggle in the title bar, embedding DeepSeek, Doubao, ChatGPT, or a custom URL as the full window via an Electron webview. Switching does not tear down either side, so a Harness session—with its scrolling and web streaming responses—can be kept running in parallel. Install with `dsh plugin add dsh-chat-shell`; state is stored in localStorage. Desktop only; depends on `dshDesktop.browser`. An optional shell patch enables persistent login but modifies the installer package. MIT licensed.
Read MoreWalmart E-commerce Video Generation and Editing: Transforming Product Assets into Deployable Final Videos
Introducing the Walmart Ecommerce Video Generation & Editing Skill on SkillHub: @infimind/walmart-ecommerce-video-generation-editing, version 1.0.0, categorized under design-media, sourced from enterprise, and published by 极睿科技-AI专家. The article starts from pain points faced by ecommerce operations, brand product teams, and livestream selling teams, outlining the skill’s capabilities: text-to-video, image-to-video, first-and-last-frame animation, reference-to-video, video-to-video, audio-to-video, AI video editing, and video extension, as well as automatic material uploading, task saving, progress querying, and final video download via AI Hive. It provides positioning, core features, activation instructions, and a typical workflow, and notes that requires_api_key=true and that specific commands and configuration fields should be subject to the official page. Suitable for AI advertising, TVCs, ecommerce videos, product showcases, livestream selling, seeding content, short dramas, comic dramas, and social media content production.
Read MoreThree.js Development: Managing 3D Web, Visualization, Animation, and Shaders with Reusable Skills
This article introduces the Agent Skill "Three.js Development" (@kunlungrowth/kg-threejs, v1.0.0, released by Kunlun Growth, categorized as dev-programming). This skill is designed to build high-performance 3D web pages, 3D visualizations, animations, and shaders using Three.js; it does not replace the framework nor does it promise instant finished products. The text begins by addressing recurring developer pain points regarding models, materials, animations, shaders, and WebGL contexts, followed by verifying practical capability boundaries. It also honestly notes data discrepancies: while the lead page indicates a rating of 99,287 and 836 downloads, the retrieved data shows a rating of 100,000, 964 downloads, and 0 installations, with the retrieved data serving as the standard. Due to failed scraping attempts on the official webpage and API, the source material lacks the content of SKILL.md, installation commands, or configuration fields. Consequently, this article does not fabricate installation steps but provides the official entry point, verification actions, and application scope, while cautioning users not to overlook mobile performance, shader positioning, and context loss.
Read MoreShort Video Viral Content Analyzer: Breaking Down Hit Videos into Replicable Creation Plans
The Short Video Viral Analyzer, released by Lingyi AI (version 1.0.3) and classified under content creation, is a video analysis and replication tool. It addresses the pain points of creators who struggle to articulate why a video went viral, how scripts are structured, or how to shoot their own content. By inputting a video file, link description, or text subtitles, the tool delivers five core capabilities: script extraction, storyboard restoration, viral cause analysis, replication version generation, and shooting execution planning, ultimately yielding actionable production plans. This article introduces the tool's positioning, core features, activation methods, typical usage, and applicable scenarios, while also highlighting considerations such as input quality and subjective judgment.
Read Moremy-first-plugin: Verifying the DSH Plugin Chain with a Single HTTP Route
my-first-plugin (LWH13345412221, v0.1.0, MIT) is a minimal web plugin example for DeepSeek Harness. After installation via `dsh plugin --profile web` and mounting in `cordis.patch.yml`, it registers a `GET /_dsh/hello` endpoint that returns JSON to verify the entire pipeline from installation to execution. The article provides the official installation command, patch snippets, curl verification steps, and directory structure, and discusses considerations such as applicable scenarios and in-process plugin permissions.
Read Moredsh-app-packager: Completing the uni-app x Packaging Pipeline in DeepSeek Harness
**dsh-app-packager** is a DeepSeek Harness workflow plugin maintained by lw0129a. It bundles the app-packager engine and CLI, enabling uni-app x project listing, environment diagnostics, iOS/Android/HarmonyOS packaging, and optional uploads within Harness sessions. It provides four agent tools: `app_packager_list`, `doctor`, `check`, and `build`, as well as an "App Packaging" sidebar panel in profiles that include a web application. The installation command is `dsh plugin --profile desktop add dsh-app-packager`, and the profile must be restarted after installation. Full functionality is available on macOS; on Windows/Linux, `check`/`build` must be run via Git Bash or WSL. Released under the MIT license; the plugin runs with DSH process privileges, so review the source code and your local security policies before use.
Read MoreIntelligent Auditing Skill for High-Risk Engineering Calculation Reports: Delegating Audit Rules and Recalculations to Scripts
Introducing the community skill "Dangerous and Major Engineering Calculation Intelligent Audit Skill" (`@user_1598e354/calculation-audit`, v1.2.1, professional, `requires_api_key=false`). This Skill audits calculation documents using a three-tier progressive structure (completeness→compliance→mechanical model). It applies a dual-code joint inspection to distinguish between the mandatory layer GB 55XXX (holding one-vote veto power), the design layer GB 500XX, and the construction layer JGJ. A `verify.py` symbolic engine independently recalculates every "satisfied" conclusion. It covers nine categories of essential documents from Document No. 48, tags normative clause numbers to issues for graded closed-loop management, and supports interaction with the Construction Plan AI Audit Skill. The article outlines capability boundaries, metadata, and verification points prior to activation. It reliably states that an official page scraping failure resulted in installation commands and configuration fields being absent from the scraped materials, and that no commands or examples are fabricated.
Read MoreStudy Abroad Skill: Aligning Student Profiles with School Lists and Application Timelines
The "Study Abroad Skill" ( @kunlungrowth/kunlun-liuxue ), developed by Kunlun Growth, is categorized under Education (version 1.1.0, source: Enterprise, requires_api_key=false). It is designed to align student backgrounds with target institutions to generate school lists, essay frameworks, timelines, and interview preparation materials, while integrating application procedures and planning knowledge into the Skill. This article outlines its four core capability boundaries, the activation method based on general SKILL.md conventions (installation commands and configuration fields are not publicly available on the official page), and usage inferred from its capabilities: first provide background, then specify goals; generate the school list before breaking down the timeline; and finally, manually verify dependencies. It also notes that the Skill does not replace the verification of admission requirements on official websites, relies on input quality, requires users to independently verify timeliness, and is currently in an early iteration stage.
Read Moredsh-design-symposium: Multi-perspective Design Seminar and Chair's Summary
dsh-design-symposium is a DSH workflow plugin (v0.5.0, MIT) maintained by Lutra23, forked from dsh-roundtable-harness. It transforms multi-model debates into multi-sage position seminars, with a chairperson outputting design meeting minutes. It supports agenda pack contracts, three-tier budgets, initial blind evaluation, fact-checking pauses, and decision journals, while providing agent tools such as symposium_start. The desktop profile requires manual link registration, and the configuration is located at $DSH_HOME/symposium.json.
Read MoreAI Lyric Creator: Lyric Writing and Style Prompt Generation Skill for Suno
# Introduction When using Suno to create songs, a common habit is to simply toss a few casually written lines into the input box. Consequently, the generated output often suffers from chaotic段落 (paragraphs) and stylistic drift; furthermore, repeated regenerations fail to resolve these structural issues. Another scenario involves developers who wish to add a "lyric writing" capability to their AI Agents but lack the time or resources to compile Suno's tag syntax and prompt formulas themselves. The "Skill" introduced below addresses these two needs by codifying lyric structure and Style Prompt formulation into a reusable skill description. # What is this? AI Lyric Creator is a skill designed for lyric writing and music generation prompts for Suno AI. Released on SkillHub, it is located at `@user_e0ebcb6a/crypto-music-pro`.
Read Moredsh-git-graph-view: Read-Only Viewing of Git Commit Graph in DSH Session Page
LUOZIR's maintained DSH community plugin, dsh-git-graph-view (v0.2.1, MIT), adds a "Git Graph" tab to the session view, providing a read-only display of the commit DAG, branches, tags, remote references, and commit details for registered workspaces, aligning with the official theme. It supports installation via both web profile CLI and Desktop GUI; it requires DSH 0.2.1 and system git, and does not write to the repository.
Read MoreWeChat Official Account Viral Articles Writing Master: A Nine-Step Writing System from Topic Selection to Retrospective
WeChat Official Account Viral Articles Writing Master (@user_557dbafb/gzh-viral-writer, v1.0.0, community source, content-creation category, no API Key required) is a writing Skill for Official Accounts that covers a nine-step workflow: material comprehension, topic selection, title hook design, narrative pacing, structural templates, style-based writing, four-level quality assurance, platform optimization, and data retrospective. It integrates Khazike-level stylistic techniques with viral-spread strategies, positioning "how to write" and "how to go viral" within a unified writing process. It is applicable to drafting, rewriting/editing, title optimization, retrospective analysis (e.g., why it failed to go viral), continuation and expansion, etc. The official documentation provides an explicit list of trigger phrases for activation. The SkillHub page shows a score of 100,000 and 915 downloads, with 0 installations. The article also covers the functional boundaries and typical use cases, and notes caveats such as installation methods depending on the runtime environment and the limited availability of community feedback samples.
Read MoreForeign Trade RFQ: Clarifying Buyer Requirements and Specification Matrices - The Organizing of Scattered Buyer Demands into a Quotable Matrix
This article introduces the community Skill "Foreign Trade RFQ Requirements Clarification & Specification Matrix" (SkillHub, business-ops, version 1.0.0, requires_api_key=false). It organizes overseas buyer emails, chat records, drawing summaries, and inquiry forms into a quotable requirements specification matrix, identifies critical gaps, and generates English clarification questions. It supports updating the RFQ ledger, version discrepancies, and quotation status based on buyer replies. It is applicable for foreign trade RFQ pre-screening and internal quotation handovers, without guessing missing specifications or replacing engineering, certification, or contract reviews. The article describes core features, general activation methods, typical workflows, applicable scenarios, and limitations.
Read MoreVoiceTranser: Local Voice Input and Auto-Paste under DSH
VoiceTranser (Lumos-789, v1.0.1, MIT) is a SkillHub client-side DSH plugin: the repository declares dsh.bundle, and after installation the agent assists with local deployment via the voicetranser skill. The backend is a macOS daemon; while holding a hotkey, it records audio over HTTP /start|/stop, transcribes locally with Qwen3-ASR MLX, SenseVoice, or mlx-whisper, corrects the text via corrections.json, and pastes it at the cursor. It is intended primarily for Apple Silicon, while Intel uses the CPU engine; Python 3.11+ and uv are required, and Accessibility permissions are used for automatic pasting. Based on the GitHub README and the SkillHub directory page, this article describes its capabilities, typical commands, and how to run it persistently with launchd.
Read MorePLC teaching plans & practical training design assistant: Leave repetitive lesson-prep tasks to Skill
This Community Skill, designed for PLC course instructional assistance, is published and maintained by user_befb1651 (Version 1.0.0). It is categorized under "education", sourced from "community", and requires no API Key. Its capabilities include lesson plan framework generation, covering progressive practical project design (ranging from start-stop control to comprehensive sorting lines), generation of accompanying documents such as assignment briefs, grading rubrics, and safety precautions, as well as built-in ideological and educational touchpoints related to the craftsman spirit and safety protocols. The article details its positioning, core capabilities, activation method (refer to the official page for details), typical conversational usage, applicable scenarios, and important guidelines: outputs are intended as preliminary drafts; safety points and grading weights must be vetted by instructors before use, and any incompatibilities with specific equipment conditions should be explicitly stated in the prompt description.
Read More