dsh-api-cost-meter: View API Costs in Real-time on the DSH Web Interface
dsh-api-cost-meter is a DSH plugin maintained by mzqmzqmzqmzq, displaying today's and the current window's API costs below the web input box, with amounts estimated in RMB based on usage recorded in session logs and official pricing. Updates are pushed by the host via SSE; the plugin reads only local logs and does not transmit conversation content externally. Installation: dsh plugin add dsh-api-cost-meter. Estimated amounts are not equivalent to platform billing statements; for models without configured pricing, only token counts are tracked.
Read More1688 Campaign Sign-up: A Skill for Delegating Platform Activity Submission to Agents
This article introduces the 1688 Activity Sign-up Skill (@daze/1688-activity-signup) released by Daze Technology. It covers activity sign-up, login, filtering, submission, and result confirmation on the 1688 platform, supporting input rules for comma-separated Activity IDs and Product IDs. Sourced from "enterance" and categorized as "business-ops", the skill is at version 1.0.0 with requires_api_key=false. The article outlines the scope of capabilities, notes that installation and activation should refer to the official page, and describes the typical submission workflow, applicable scenarios, and precautions.
Read Moredsh-kb: Consolidating Session Knowledge into Local Immutable Fact Cards
dsh-kb (MyRemme) is a memory plugin for DeepSeek Harness: it uses SQLite fact cards as the single source of truth, employing skeleton-tree classification, relationship graph evolution, and FTS5 Chinese bigram retrieval with zero external dependencies. It supports tools such as kb_search, kb_add, and kb_ingest, as well as a sidebar and CLI. Drawing on the official README, this article covers installation (pnpm add dsh-kb), configuration, typical commands, and the division of labor between ingest and conversation extraction, and explains the design orientation that prioritizes addressing distortion and decay over pure recall.
Read MoreLiu Run, Business Thinking Consultant: Delegating Business Analysis to a Systematic Thinking Framework
Lius Thinking Advisor is a community Agent Skill based on the SKILL.md format, maintained by user_b0de6696, version 1.0.0, categorized as professional, and does not require an API key for use. It integrates Liu Runs business thinking framework into the Agents analysis workflow. Its core tools include the Five-Dimensional Thinking Method, Five Thinking Building Blocks, and Foundational Logic. It is suitable for business model analysis, strategic decision-making, problem diagnosis, trend assessment, and retrospective summarization—scenarios requiring deep, systematic business analysis—and is explicitly not suitable for simple factual queries or purely computational tasks. This article covers its positioning, core features, official installation address, typical usage, and applicable limitations. For dynamic ratings and download data, please refer to the official page.
Read Moredsh-conversation-folders: Assign dedicated folders and auto-organize DSH conversations
Introducing the DSH plugin `dsh-conversation-folders` (v0.4.0, MIT) maintained by `myqy91615-web`: It creates title-named folders for conversations with sent messages within the workspace, redirects the working directory to write conversation logs to `.md` files, and automatically classifies idle (default 10 minutes) conversations by file extension with a research sequence number prefix. The menu allows for the deletion of conversations and their content. Includes official GitHub installation commands, `cordis.patch.yml` configuration items, verification steps, and API documentation.
Read Moredsh-restart: Add a "Restart" button to the DeepSeek Harness Desktop title bar
dsh-restart is a community client plugin that adds a "Restart" button next to the "Application/Edit" option in the title bar of DeepSeek Harness Desktop (Windows only). Clicking the button only pops up a confirmation dialog. After confirmation, it terminates the Electron main process and Host through the Host and detached assistant, then launches the executable file again, which is equivalent to exiting and reopening. The Host requires a POST request to /dsh-restart/restart?force=1; otherwise, it returns 409. The installation command is `dsh plugin --profile desktop add github:mwk719/dsh-restart`, which requires dsh>=0.2.0-rc.1, and the log is located at ~/.dsh/dsh-restart.log.
Read Moredsh-mote-media: Jimeng-style mote image and video generation in DSH
Maintained by MuRanSea, the `dsh-mote-media` DSH bundle plugin facilitates multi-path image and video generation via the unified `mote` channel and OpenAI-compatible image channels. It offers a "Jimeng-style" web creation station, the `mote-image` chat skill, and four categories of tools: `mote_models`, `mote_image`, `mote_video`, and `mote_media_task`. Configuration is defined in `cordis.patch.yml` or the creation station's channel settings, and it can be installed via the plugin page's local directory or using `plugin_manager install_bundle`. It is ideal for centralized management of keys, tasks, and local assets within the Harness.
Read MorePhysics in Motion - University Physics: Transforming University Physics Simulation Pages into Direct Access Links for Knowledge Points
"Animated Physics Learning - University Physics" (@yimubiao/university-physics-lab, Version 1.0.2, Source: enterprise, Category: education, requires_api_key=false) is a simulation experiment Skill designed for university science and engineering physics, published by Yi Mubiao. Its core behavior follows a single rule: when a user mentions knowledge points such as the Lorenz Attractor, Double Pendulum Chaos, the N-body Problem, the Kinetic Theory of Ideal Gases, Black-body Radiation, Vortex Ring Evolution, Wave Interference and Diffraction, Wave Propagation, and Simple Harmonic and Damped Oscillation (including synonyms and colloquial terms listed officially), it directly returns hyperlinks to simulation or interactive pages that can be opened online with a "?kp=" parameter. If no match is found, it returns a link to the Skill's homepage. The output consists of links rather than explanatory text and requires no API keys or local simulation environments. This article outlines the matching and output mechanisms, the list of covered knowledge points, verifiable enabling information, and two typical types of invocation results. It also specifies applicable scenarios and limitations: coverage is limited, output does not include explanations of principles, and the provided documentation lacks installation commands and specific "?kp" values. Therefore, users must verify these details item by item via the official SkillHub page before integration.
Read MoreMulti-Platform Text Refresh Workshop: Rewriting One Article into Platform-Ready Versions
Introducing a community Skill: rewrite articles or paragraphs into highly original versions without altering the original meaning or violating compliance baselines. It features five built-in modules: Basic Paraphrasing, Deep Restructuring, Style Transition, Scenario Adaptation, and Compliance Check, along with a pre-rewrite risk pre-scan via `preflight.py`. It automatically determines the rewriting depth and presentation format based on platform attributes for channels such as WeChat Official Accounts, Toutiao, Zhihu, Xiaohongshu, Douyin, Weibo, Bilibigli, Channels, Baijiahao, NetEase, Sohu, Xueqiu, and Douban. When no specific platform is specified, it defaults to generating a Deep Restructuring version for WeChat Official Accounts. It is marked as `requires_api_key=false`; please refer to the main text for the official address.
Read MorePhysics Learning Assistant: University Physics Formula Quick Reference and Problem-Solving Step Guide
Physics Learning Assistant is a SkillHub community skill located at @user_3b5a77fc/physics-quickref. Version 1.0.0, categorized under education, was published by community developer user_3b5a77fc. It targets formulas, symbol definitions, direction determination rules, and the retrieval of common pitfalls in university physics studies, covering three major modules: electrostatic fields, electromagnetism, and mechanical waves. Users can directly ask "What is the XX Law?", "How to determine the direction of XX?", or "What is the formula for XX?". The skill returns corresponding formulas, symbol meanings, rule explanations, and common traps in multiple-choice questions, while assisting in organizing problem-solving steps. The tag `requires_api_key=false` indicates that it operates without external API keys. The article provides the official address and SkillHub page, explaining core functions, reproducible query examples, target audiences, and limitations: it serves only as a retrieval and problem-solving aid, does not replace textbook derivations, and does not cover content outside the three specified modules. Sourced from the community, the page shows a rating of 100,000, 895 downloads, and 0 installations; it is recommended to conduct a small-scale validation before team use.
Read Moredsh-grok-stars: Adding Grok-Style Starry Background to DSH New Session Page
dsh-grok-stars is a DeepSeek Harness Web client plugin that overlays a Grok-style animated starfield on the background of new session tabs, inspired by grok.com. The installation command is `dsh plugin --profile desktop add dsh-grok-stars`. Version 1.1.0, MIT licensed, integrates through cordis.patch.yml and client injection. The documentation does not provide additional configuration options, making it suitable for DSH users seeking a lightweight and fun customization. Please review the source code before installation, and note that the plugin runs with dsh process privileges.
Read Moredsh-webui-auth: Adding Persistent Authentication Gate to DSH WebUI
@multidsh/dsh-webui-auth, maintained by multidsh, provides disk-persisted sessions and four-layer route wrapping (static assets, /plugins, /api, and WebSocket) for DeepSeek Harness WebUI, blocking access when unauthenticated. It modifies no core source code and enforces a fail-closed state upon enabling failure. Forked from dsh-webui-auth@0.3.6, it incorporates the 0.3.7 fix regarding login redirects with the X-Forwarded-Host header in reverse proxies. The article details npm/GitHub installation commands, deployment considerations for trustedHosts and public-url, as well as adaptations for the DSH Desktop profile, host bridging, and UI-level locking. Review the MIT-licensed source code and the implications regarding process permissions before installing.
Read MoreInternet Workplace Language Master: A Skill Covering Four Communication Scenarios
“互联网职场语言大师” is a community-sourced WorkBuddy Skill (version 1.0.0, category: content-creation) that integrates four high-frequency workplace communication needs: Script Transformer (five tone versions—upward reporting, polite cross-department communication, tactful refusal, progress chasing, and anti-PUA—with support for multiple versions at once); OKR/Performance Polisher (packaged as Objective + Key Results, supporting IC/TL perspectives and self-review, peer review, and work-reporting scenarios); Internet Jargon Translator (bidirectional forward and reverse translation, adjustable jargon density in three levels, and assigning each text a 0–100 jargon index); and Instant Email Reply (automatically identifying incoming message types such as progress chasing, help requests, complaints/feedback, and supporting five tones—formal/casual/firm/delayed/proactively cooperative—adapted to WeCom, email, and TAPD). No API key is required. The article provides official examples of jargon translation and script conversion, and notes that the fetched materials contain no installation command and that the official page is inaccessible, so installation steps were not fabricated.
Read MoreHeyGen Video Generation Alternative | AI Video Generation and Editing: One Skill Covers Text-to-Video
This Skill integrates AI video generation and editing into a single workflow: users can either start from scratch with text inputs or utilize images, videos, or audio to control the output. When used with AI Hive, required assets are automatically uploaded prior to generation, and once submitted, tasks are saved automatically with progress tracking and final video downloads, requiring only the user's requirements and assets. Official capabilities cover 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. These capabilities are applied to AI advertising, TVCs, e-commerce videos, product demonstrations, live sales, lifestyle recommendations, short dramas, comic dramas, and social media content. Released by Jirui Technology - AI Experts, located at @infimind/heygen-video-generation-alternative. Version 1.0.0. Category: design-media. Tag: requires_api_key=true; an API Key must be prepared before use. As the official documentation does not provide specific installation commands or environmental differences for Cursor, Codex CLI, or Claude Code, this article makes no speculations; please refer to the official page for access.
Read MoreShould We Go Global? Overseas Expansion Readiness Assessment (5-Dimensional Scorecard Board Report)
This article introduces the community Skill on SkillHub: "Should You Go Global? Go-Global Readiness Checkup (Five-Dimensional Scorecard & Board Report)." Published by @user_69f3cf82 (Version 2.2.0) and categorized under business-ops, it currently holds a rating of 100000 with 832 downloads. Marked as requires_api_key=false, it can be enabled without an API key. It conducts a deep diagnostic across five dimensions: motivation, pace, risk, organization, and judgment. It outputs a tri-option conclusion (Ready to Go / Hold Off / Discuss Once Conditions Mature), a five-dimensional scorecard, red-flagged fatal flaws, scenario projections (Worst / Neutral / Best case), an action roadmap with responsible parties and deadlines, as well as information gaps and confidence levels. All figures must be traceable; unverifiable data is marked as "To Be Supplemented," refusing to substitute estimates for actual measurements. The final decision-making power remains with the user. Trigger scenarios are limited to: whether to go global, ability to go global, go-global assessment, go-global readiness, overseas market judgment, and go-global decision-making. The checkup is not performed in four specific cases, which should be redirected to execution-focused tools or handled as Q&A: users who have already decided to go global and only need to select countries/tracks/landing resources, pure information or academic inquiries, non-operator entities, and those simply comparing prices for managed service providers. Based on the official introduction, the article restores the capability boundaries of the seven modules, a reproducible six-step typical usage flow, suitable and unsuitable user groups, and three key considerations regarding data quality, data supplementation responsibilities, and self-determination boundaries.
Read MoreMulqi ReaSLL: Loading the "Inverse Reconstruction + Self-Learning Closed-Loop" Workflow Skill in DSH
Mulqi ReaSLL (mulqi-reasll 1.0.2) packages Speedrunlab's reverse-engineer-anything-self-learning-loop as a DSH workflow plugin. It introduces no new tools, registering only the `reverse-engineer-anything-self-learning-loop` skill. The workflow consists of audit, 100% checklist, implementation, evidence review, and gap iteration until success. Requires DSH >= 0.2.1-rc.2 and Node >= 22.19.0; Section 0 intake is optional starting from 1.0.2. Install via `dsh plugin --profile web add github:mulqizamzam/mulqi--reasll`; takes effect in new sessions. MIT licensed, maintained by mulqizamzam.
Read MoreTour Guide Qualification Practice Assistant: A Skill for Tour Guide Certificate Exam Question Practice
"Tour Guide Qualification Exam Practice Assistant" is a community-sourced Agent Skill on SkillHub (identifier: @user_af163eaa/z-brush-guide-cert, version 1.0.0), categorized under education, published by user_af163eaa, and requires no API key. Tailored for exam preparation, it features practice questions generated by AI based on a general syllabus. Explicitly noted as unaffiliated with official exam predictions, its content is strictly confined to five modules: tourism policies and regulations, tour guide practices, national tour guide fundamentals, local tour guide fundamentals, and oral examinations. This article outlines its positioning, verified capabilities, installation and activation boundaries, typical use cases, and precautions. Due to a failed fetch of the official page, the installation commands and configuration parameters remain unverified; therefore, this article provides no unverified commands and makes no differential assumptions regarding Cursor, Codex CLI, and Claude Code.
Read Moredsh-translation-plugin: Long-form Translation and Polishing within DSH
Mulibaixi's dsh-translation-plugin provides a translation and polishing workspace for DeepSeek Harness, supporting Quick short translation and Normal mode with main model collaboration, topic memory, version retention, and local history. Installation requires obtaining the dsh-translation-plugin .tgz from GitHub Releases and enabling it in the plugin manager; Normal mode requires dsh-composer-extension 0.4.5 to be installed first. It is suitable for long translation projects such as novels, game scripts, and academic works that require consistent terminology across sessions and multiple rounds of review.
Read Moredsh-composer-extension: Linked Submission and Task Collaboration between Primary Chat and Translation Workspace
Maintained by Mulibaixi, the dsh-composer-extension is a DSH workflow plugin that bridges the main chat with the dsh-translation-plugin workspace. It supports submitting instructions alongside materials, passing aggregated context and requirements from the main model to the translation model, integrating notifications for critical task stages, and handling messages for stopped and expired tasks. Installation involves using the plugin manager to specify the path or URL for the dsh-composer-extension version tgz; it requires DSH >=0.2.0-rc.2 and the companion translation plugin. Automatic maintenance of host patches is available for Windows Desktop. Licensed under MIT, the current documentation version is 0.4.5.
Read MoreDesign of Biomedical Experimental Protocols: A Standard Workflow from Literature Anchoring to Experimental Design
# Introduction When drafting experimental protocols for biopharmaceutical research service companies, a common practice is to have an AI model generate the experimental steps first and retroactively append a few references. The issue with this approach is that controls, dosages, and detection endpoints often lack literature support, the origins of innovative points remain unclear, and the rationale for specific steps is uncertain when clients inquire. This Skill establishes a fixed sequence: conduct literature retrieval first, followed by experimental design, ensuring that every key decision in the protocol corresponds to a verifiable article. # What is this **One-sentence positioning:** Given a broad research direction, the system automatically selects one recent high-impact review and three high-impact experimental design articles to provide the overall framework, and subsequently generates an experimental protocol based on these findings. **Publisher and Maintenance:** Published by user_69bc6565; Source: Community; Classification: Professio...
Read MoreMuseScore-MCP: Searching MuseScore Scores and Reading Metadata in MCP
musescore-mcp is an MCP Skill designed for searching MuseScore scores and reading metadata. It is published on SkillHub at @user_15292d5a/yjkj-musescore-mcp, version 1.0.0, source community, category knowledge-management, tag requires_api_key=false. The page shows a rating of 100000, 875 downloads, and 0 installations. It provides two functions: searching MuseScore scores by criteria and retrieving metadata for a specific score, with typical fields such as license type and page count. It activates for any request involving MuseScore scores, arrangements, or sheet-music metadata. Official trigger phrases include: "find sheet music for", "search MuseScore for", "is there a free arrangement of", "what's the license on this MuseScore score", "how many pages is". Before enabling, three hard prerequisites must be met: musescore-mcp is installed, the fetchproxy extension is active, and there is a logged-in musescore.com browser tab; none can be omitted, as it uses the authenticated tab as a scraping channel rather than relying on a public API. This article only states verified facts: the official Setup section could not be retrieved during this fetch, so installation commands, directory paths, and configuration fields are not included, nor were client-specific instructions fabricated. It is suitable for users who need to search scores within conversations or automation workflows and then use license and page count as intermediate results, such as cataloging a library or finding reference sheets before arranging; it does not perform score editing or generate notes.
Read MoreMuciya DSH macOS Suite: Five Independently Installable DSH Web Client Plugins
Maintained by MuCIYa953 and released under the Apache-2.0 license, this suite splits “DSH Web Suite for macOS” into five independent bundles—plugin market, skills gallery, look gallery, agent team, and usage dashboard—each instantiating the same cordis and client objects on each. This article walks through the responsibility of every bundle, DSH ≥ 0.2.0-rc.2 and macOS-side toolchain prerequisites, official Release or sub-path installation commands, common start-up and verification steps, and a quick troubleshooting checklist. 1. Packages and responsibilities • plugin-market: Browses, installs, and upgrades third-party plugins; exposes the same hook protocol as DSH core. • skills-gallery: Hosts reusable skill cards and one-click import flow with auto-schema validation. • look-gallery: Supplies themes, icon sets, animation packs; hot-reload supported. • agent-team: Manages multi-agent rosters, quotas, and A/B rollouts; spawns isolated cordis namespaces. • usage-dashboard: Renders real-time token, cache, and billing metrics via WebSocket; can run directly in the browser. 2. Runtime dependencies • DSH >= 0.2.0-rc.2 (minimum DSH core version for the new plugin spec) • macOS 12+, recommended Node.js >= 18.4, npm >= 9. • Xcode Command Line Tools for node-gyp rebuilds when native modules are involved. 3. Installation 3.1 One-click via Release ``` curl -L https://github.com/MuCIYa953/dsh-web-suite/releases/latest/download/Install.sh | bash ``` 3.2 Manual sub-path deploy ``` git clone --depth 1 https://github.com/MuCIYa953/dsh-web-suite.git ~/dsh-web cd ~/dsh-web && npm ci npm run dshbundle -- --all --prefix /usr/local/lib/dsh ``` Both routes build `bundles/{plugin,skills,look,agent,usage}/*.bundle`, which can later be mounted with `dsh attach`. 4. Launch, verification, debugging 4.1 Start ``` dsh attach /usr/local/lib/dsh/plugin-market.bundle dsh attach /usr/local/lib/dsh/skills-gallery.bundle dsh attach /usr/local/lib/dsh/look-gallery.bundle dsh attach /usr/local/lib/dsh/agent-team.bundle dsh attach /usr/local/lib/dsh/usage-dashboard.bundle dsh start --port 8022 --tls --inspect ``` 4.2 Verify ``` curl -I https://localhost:8022/~plugin-market curl -I https://localhost:8022/~usage-dashboard ``` 200 responses and visible charts mark success. 4.3 Debug & metrics ``` # Node inspector on port 9229; trace WebSocket traffic dsh logs --plugin plugin-market safari -o "https://localhost:8022" -H "X-DSH-Debug: 1" ``` 5. Troubleshooting • 502 loop: ensure DSH ≥ 0.2.0-rc.2 and that /usr/local/lib/dsh contains all five bundles. • self-signed cert warnings: bypass via `curl -k` or add the CA to Keychain. • bundles not found: re-run `npm run dshbundle` and confirm `~/dsh-web/dist/bundles` is complete. • permission errors: prefix install/start commands with `sudo` or grant Full Disk Access to Terminal under “Privacy & Security”. 6. Release notes v1.0.0-alpha: first RC; gas-friendly WebSocket heartbeat in dashboard; bug fixes in agent-team re-election logic.
Read MoreExpense Reimbursement Capability: Structuring Invoices, Receipts, and Key Points of Expense Approval
This article introduces the "Reimbursement" Skill on SkillHub (released by Kunlun Growth): for enterprise financial automation purposes, covering invoice/document organization, compliance checks, reimbursement claim generation, and approval point extraction. Section II: Skill Positioning. The article outlines its positioning, verified core capabilities, public metadata, confirmable and unconfirmable installation and activation details, typical use-case boundaries, and applicability considerations, and provides the official download link and the SkillHub page.
Read MoreRequirement-Driven Travel Planner: Directly Converting Travel Requirements into HTML Itineraries
This is a domestic multi-day itinerary Agent Skill located at `@user_e0c9a7b3/travel-itinerary-planner-jyj`. Published by `user_e0c9a7b3` from the "community" source under the "life-service" category (Version 1.0.0), the page is marked with `requires_api_key=true`. It treats seven dimensions—origin, transportation, time, dining, accommodation, play preferences, and budget—as constraints. It conducts research across candidate libraries by dimension: utilizing Amap's street-scanning rankings for dining, Xiaohongshu for attractions, and price comparisons across Ctrip, Qunar, and Tongcheng for accommodation. It then generates a default route and renders it into a single HTML file containing Tencent Map numbered markers, a self-driving duration matrix, and floating guide cards. The article explains the skill's positioning and four-step capabilities, access points, and API Key configuration considerations. It outlines a reproducible workflow: inputting data across the seven dimensions, confirming the candidate set, checking three types of structures, and adjusting the route. Additionally, it highlights caveats regarding the platform attributes of the candidate libraries, the scope of accommodation price comparisons, and the coverage boundaries of self-driving. It also warns that metrics such as a rating of 100,000, 919 downloads, and 0 installs should not be interpreted as quality endorsements.
Read MoreDSH-MSSZSD-Toolkit: A Collection of DSH Workflow Utilities
The dsh-msszsd-toolkit (v0.5.0, MIT), maintained by MSSZSD3925, consolidates QQ remote driving, cross-session questioning, interface pausing, restart-and-resume, and workday peak-hour discipline injection into a single DSH plugin, with the detail page providing runtime toggles for five modules. Based on the official README, the article covers key points including npm/GitHub installation commands, cordis group configuration, local JSON overrides, and QQ command and approval mirroring, while cautioning about bundles verification and same-process permission risks.
Read More