Task Deck: Multi-Session Task Console on DSH Desktop

Task Deck (astephengoo-arch/dsh-task-deck) is a DSH Desktop client plugin that reads real-time session directories, presents each conversation as a task card, and provides a sidebar dashboard, details panel, Task Thread, Continue/Fork/Handoff functionality, and a personal localStorage annotation layer. It supports automatic "Done" status when idle, Blocked/Review badges, and a loopback-only stats interface (including token costs and cost dimensions based on an editable pricing table). Installation is accomplished by loading Release tgz or source code directories via `plugin_manager install_bundle`; requires DSH >= 0.2.1-rc.5 and a Web/Desktop client with sidebarRight support. MIT licensed, does not copy transcripts, and performs no external link analysis.

Read More
dsh-status-balance: Session Spending and Wallet Balance in the Web UI Status Bar

ash ai’s DSH client plugin, `dsh-status-balance`, permanently displays the current session cost and wallet balance below the Web UI input box; clicking allows users to view today's consumption, balance deduction, estimated days of availability, and peak/valley periods. It is zero-dependency and requires no build; it does not enter the conversation chain or consume extra tokens. The Host fetches and caches the balance, while the browser polls visibly every 60 seconds. Installation requires the web profile; the command is `dsh plugin --profile web add github:ash ai/dsh-status-balance`. Amounts are estimates based on official pricing, not actual invoices; please review the source code and MIT license before use.

Read More
dsh-effort-slider: Replace the DSH editor's reasoning levels with an animated three-tier slider

dsh-effort-slider is a DSH client plugin maintained by asakumizy. It takes over the `conversation.input.model` slot in the Composer input area, replacing the native model and reasoning level picker with a three-tier slider offering Medium, High, and Very High. It supports dragging, track clicking, and keyboard interaction, committing the selection on release. Switching does not trigger reordering or dimmed states, and the control features a spring-driven thumb and particle fill animation. When model catalog parsing fails, it retries and displays the failure reason on the button. To install, run `pnpm add` with a local path under the profile and mount `@local/dsh-effort-slider` in `cordis.patch.yml`. Released under the MIT license.

Read More
AI Grader for New Gaokao English Composition: Free Scoring and Deduction Location

New Gaokao English Essay Smart Grader (Free Trial Version), published by @user_a7cd9e27 on SkillHub (Version 1.0.1, Category: education, No API Key required). Users can paste their New Gaokao English essays (15-point applied writing or 25-point continuation writing). The free version provides a suggested score and grade level, three main reasons for point deductions, and a one-sentence overall review, facilitating quick score estimation and identification of deduction points. Features such as sentence-by-sentence correction, upgraded rewriting, reference model essays, and teacher/batch grading are available in the paid version, gaokao-essay-grader-pro (¥1.00 per session). Trigger scenarios include grading English essays, grading applied writing, grading continuation writing, identifying essay deductions, evaluating essay scores, and improving Gaokao English essay performance.

Read More
dsh-genui-enhanced: Rendering Interactive GenUI in DSH Conversations

Asaiuta/dsh-genui-enhanced is a local fork (0.14.0-fork.3) of @changfenhuang/dsh-genui 0.14.0 that renders interactive UIs within DeepSeek Harness responses via dsh-ui fences. Compared to upstream: diagrams use pele (smaller resources, CSS variable themes), chart areas, native date/number inputs, table pagination, and Computed fields; it is not published to npm and requires building with pnpm followed by registration via link. This article describes environment requirements based on the README, installation differences from the npm upstream, and 60-second verification steps.

Read More
WorkBuddy Usage Dashboard & Floating Orb: Local Parsing of Session Records for Credit and Token Usage Insights

Although the local WorkBuddy lacks a backend for tracking credit or token consumption, session logs are stored in `~/.workbuddy/projects/*/*.jsonl`. This Skill directly parses local JSONL records to provide a credit/token consumption dashboard, comparisons of the most expensive conversations, and usage distribution across models and projects. It also includes a macOS desktop floating ball that displays the number of active conversations and today's consumption in real-time. It requires no API keys, operates offline, keeps data local, and utilizes a pure standard library with zero dependencies for the dashboard. This article introduces what it is, its capabilities, installation and activation instructions, typical usage scenarios, and limitations.

Read More
dsh-forge: Retrieval and Construction Governance in DSH

dsh-forge is a workflow plugin developed by Asaiuta for DeepSeek Harness, integrating fff Rust search, Tree-sitter symbols, and an editing engine with ledger snapshots. It provides three default search tools—ff_find, ff_grep, and ff_symbols—while enhancing the built-in read, grep, and glob functions. When the surface is set to 'full', 10 tools for editing, rollback, Trace, and project management become available. It has not yet been published to npm; it requires Node ^22.19||>=24 and DSH >=0.2.0-rc.2, and can be installed via profile link. Data is stored in ~/.dsh/forge, supporting rollback even without Git.

Read More
dsh-better-model-selector: Align inference tiers and context length with the official catalog

The DSH community plugin dsh-better-model-selector (Asaiuta) reads the official model directory within the installation package, writes the correct inference tier for each model, and adds a "context length" row to the model selector (default / 262K / 524K / 1M / 2.1M). It performs no endpoint probing or network requests. To install, use pnpm in the desktop profile to add it from GitHub, write it into bundles, then restart. An optional cordis.patch.yml allows configuration of mode, capacity, and other settings. It is suitable for users of custom gateways; note that official specifications do not necessarily equal gateway capability, so please review the source code and MIT license before installing.

Read More
WPS/Office Full-Pipeline Analysis: 12 Commands for docx/PDF Document Processing

**# Introduction** Developers working on document processing frequently encounter the following requirements: splitting exam paper .docx files into individual files by question number, merging bid documents into a single volume, checking dozens of Word files for duplication, and encrypting PDFs or converting them back to Word. Finding separate scripts or libraries for each requirement involves dealing with different environments, formats, and parameters, making workflow integration tedious. The Skill introduced below consolidates these operations into a set of unified commands. **# What is this** The WPS/Office Full-Link Parsing Skill is a community Skill on SkillHub. It is located at @user_638e256a/doc-processor-v2. Currently at version 1.3.0, it is released and maintained by user_638e256a and categorized under office-efficienc...

Read More
Hummingbird AI Skill: Delegating Image Translation, Video Translation, and AI Image Generation to Agents

Fengniao AI is a community Skill on SkillHub (@user_d7c8a17c/fengniaoai, version 1.5.1, categorized under design-media, requiring an API Key). Designed for e-commerce and content marketing scenarios, it offers three categories of capabilities: image translation, video translation, AI image generation, AI retouching, and AI video processing. The article covers its positioning, keyword trigger mechanisms, use cases, prerequisites, and limitations.

Read More
Introduction to the Bid Document Writing Expert Skill

"Bid Document Creation Expert" is a Skill published by community author user_37b3d0df, located at @user_37b3d0df/biddocx-creator, version 1.1.0, categorized under professional, sourced from community, and tagged with requires_api_key=false. It targets the scenario of writing technical proposals for bidding, integrating requirement understanding, solution design, functional design, architecture diagramming, compliance response, quality inspection, and packaging into a single workflow, covering informatization projects in government, IT application innovation, enterprise, healthcare, and education sectors. The article introduces the Skill's positioning, six core capabilities, trigger mechanism (no API key required, triggered via natural language), four typical use cases (generating proposal frameworks, compliance checking, generating standalone diagrams, and organizing bid materials), as well as usage limitations: community-sourced Skills require reading through SKILL.md first; the page displays 100,000 ratings, 1,101 downloads, and 0 installations; manual review remains necessary before final packaging.

Read More
File Annotator (dsh-plugin-file-annotate): Annotate selected text in the DSH right sidebar and submit to the conversation draft

File Annotator (Arvin-work/dsh-plugin-file-annotate) is a DeepSeek Harness client plugin that opens workspace text and code via a lazy-loaded file tree in the right sidebar. It allows users to write inline annotations on selected text, supports unified 1/N navigation within sessions, and offers Markdown preview. Upon submission, annotations are appended to the conversation input box without auto-sending. The plugin has no build-time or runtime npm dependencies and is released under the MIT license. To install, run: `dsh plugin --profile <profile> add github:Arvin-work/dsh-plugin-file-annotate`. Annotations are stored in localStorage, with read and write operations handled through the host workspace API.

Read More
dsh-public-edge: DSH Public Network Access and LAN Bridge

dsh-public-edge is a DSH networking plugin maintained by ArtificialAngels. It provides one-click management of Cloudflare Quick Tunnels via the settings page, offering public HTTPS links with tokens and QR codes. It also includes a built-in port 8787 LAN bridge (Host rewriting, Origin stripping, and WebSocket pass-through). The plugin automatically downloads `cloudflared` if it is missing, and configuration can be adjusted for ports and mirrors via `cordis.patch.yml`. Since it is not listed on npm, it must be installed from GitHub followed by a DSH restart. Users must treat the token as equivalent to local machine control authority and be aware of Cloudflare's plaintext transmission and Quick Tunnel limitations.

Read More
Comprehensive Lyric Creation: Delegating the Lyrical Process from Topic Selection to Refinement to Skill

This article introduces the premium Skill on SkillHub, 'Ultimate Lyric Creation' (@ixhlink/ixhlink-skills-lyric-gen, v1.0.1, powered by Xinghua Zhilian Technology). It breaks down lyric writing into actionable steps such as topic selection, sensory mining, metaphor construction, structural development, rhyme design, and refinement to eliminate AI traces. Integrating Pat Pattison's sensory writing method from Berklee, collision techniques for metaphors, and the Chinese 'Shisantie' (thirteen zhe) rhyme system, it supports various musical genres including pop, folk, rap, GuFeng, rock, electronic, and R&B. The article outlines its core capabilities, trigger words, and metadata (which requires payment but no API Key), and points out that since the official website does not provide installation commands or examples, usage should strictly refer to the official page.

Read More
Knowledge Extraction Skill: Organizing Raw Content into Structured Concise Notes

The "Knowledge Extraction" skill is a community-sourced, generic SKILL.md Skill located at @user_285cce02/knowledge-extraction. It is version 1.0.0 and categorized under knowledge-management. It provides a unified specification for transforming raw content (ASR transcripts, OCR text, articles, video transcripts) into structured, refined Markdown notes. The pipeline includes: summary → outline → three-level nested body → knowledge distillation → linked notes → keywords, and it also includes formatting conventions and a quality-check checklist. The tag requires_api_key=false indicates that no additional API key is needed. The official page failed to be retrieved this time; therefore, the installation command, configuration fields, and complete examples are unverified. No content has been fabricated in the body. It is recommended to refer to the official address and validate the output quality with sample data before deployment.

Read More
WeChat Official Account Layout Engine Pro 3.0 (Illustrated Version): Content-Driven Chinese WeChat OA Layout Skill

This is a paid formatting Skill for Chinese WeChat Official Account articles, released on SkillHub by Beijing Zixuan Ren Yi Technology Service Center (Version 1.4.0, Content-Creation). It generates exquisite, content-driven layouts supporting five visual directions and custom styles. Users can first preview the layout on a real mobile first screen for free. Full-text generation and optional image generation are delivered only after real-time price confirmation, payment, and online review of the original order, providing secure rich text, a copy workbench, and image assets. This article outlines its positioning, core features, typical workflow, and usage notes.

Read More
dsh-plugin-midclick-close-tabs: Add middle-click-to-close for DeepSeek Harness right sidebar tabs

This document introduces the DSH client plugin "dsh-plugin-midclick-close-tabs", maintained by aRooooooba. On the web, it enables closing tabs in the right sidebar (e.g., files, terminal, browser, previews) and on floating window title bars via the middle mouse button. By triggering the built-in close button, it preserves existing rules such as terminal recycling and maintaining the last navigation tab. It also explains the SkillHub catalog information, the installation and uninstallation commands in the README, the commit-pinned installation plan, typical usage without configuration, and security considerations.

Read More
Topic Selection: The Skill of Filtering High-Value Topics Using a Four-Dimensional Evaluation Method

"选题" is a Skill published on SkillHub, available under the handle @kunlungrowth/kunlun-xuanti. Released by Kunlun Growth as version 1.1.0, it falls under the content-creation category. It evaluates candidate topics along four dimensions—demand, differentiation, feasibility, and monetization—and outputs a list of selected topics along with the reasoning behind each item's selection or rejection. It covers use cases including content creation, course development, and event planning. This article introduces its positioning, core functionality, usage methods, applicable scenarios, and limitations.

Read More
Article AI Tone Detection and Humanization Tool: A Skill for Detecting and Rewriting Chinese Content to Remove AI Characteristics

ai-text-editing is a community Skill dedicated to Chinese AI text detection and rewriting (published by user_c71188e3, v1.0.0, categorized under content-creation). It employs a four-dimensional quantitative engine to scan text and output an "AI-flavor" score from 0 to 100. The lexical layer covers five categories of blacklisted words and provides replacement suggestions; the syntactic layer identifies 15 types of template sentence patterns; the structural layer measures naturalness using connective density, sentence length standard deviation, and paragraph variation coefficients; and the emotional layer detects formulaic expressions such as moralizing conclusions. The rewriting process follows a progressive five-step workflow: detection and localization, lexical replacement, sentence restructuring, structural rhythm optimization, and the injection of genuine emotion. It offers differentiated strategies for various scenarios (academic papers, new media, business reports, creative writing, and email notifications) and is equipped with specialized high-risk vocabulary databases for disciplines such as medicine, law, and education. The detection script supports command-line invocation and JSON output, allowing integration into automated content pipelines. The tag `requires_api_key=false` indicates that no external API key is needed. As specific installation commands and parameter fields are not provided in the materials, please refer to the official page for usage instructions.

Read More
Finance and Tax Test: Integrating the Finance and Tax Practical Training Knowledge Base into Conversational Q&A

Finance and Tax test is a community-sourced Agent Skill, located at @user_92b23333/caishui-cathy, version 1.0.0, category professional, published by user_92b23333. It connects directly to the Tencent Cloud ADP Tax and Finance Training Knowledge Q&A Application, answering questions related to reimbursement, invoice auditing, tax benefits, and classroom cases based on the configured finance and tax knowledge base. The tag requires_api_key=false indicates that no separate API Key is required. The article introduces its positioning, core functions, installation and activation methods, typical question examples, as well as applicable scenarios and precautions: answers depend on the coverage scope of the knowledge base, with a current installation count of 0, and it is recommended to validate it with classroom cases before integration.

Read More
Litigation Case Review and Case Handling Summary: Transforming Individual Case Processes into Law Firm Knowledge

A case review and knowledge consolidation tool designed for litigation teams: input the complaint, judgment, and representation statement to generate court transcripts, case summaries, in-depth review reports, extraction of case handling experience, and knowledge cards via a four-mode intelligent review system. Released by user_281f816d, from the community, categorized under "professional", version 1.0.0, tagged as requires_api_key=false, which means no separate API Key application is needed. Upon verification, both the official API and the SkillHub page are inaccessible. The provided materials contain no installation commands, configuration fields, or reproducible examples; therefore, the capabilities and workflows are described solely based on the official brief, and unverified installation steps are not provided. Its application prerequisite is the availability of fully drafted litigation documents; specific differences among the four modes are not elaborated by the official sources. The page indicates zero installations. It is recommended to test with non-critical cases first before promoting the tool, after confirming the quality of its outputs.

Read More
Intraday and Short-to-Medium-Term AI Trading Strategy Analyst: Codifying A-Share Entry/Exit Analysis into an Executable Process

This article introduces an A-share intraday and mid-short-term buy/sell point analysis Skill published by community developer user_d4779d66: it automatically identifies individual stocks/ETFs and T+0/T+1 trading mechanisms, calls the westock-mcp / wind-finance connectors to retrieve data, performs pre-filtering based on liquidity, market conditions, sectors, events, and price limit thresholds, and then outputs signal strength, hard stop-loss levels, profit targets, position mapping, and next-day contingency plans according to a signal rule library. The article notes that the installation command and output samples are missing from the retrieved materials and are not fabricated; it also provides three reproducible example prompts, applicable scenarios, and key points of the official disclaimer.

Read More
All-in-One Foreign Trade Assistant: A One-Stop Multilingual Workbench for 31 Trade Scenarios

The Foreign Trade All-in-One Assistant is a foreign trade business Skill published on SkillHub, covering 31 foreign trade scenarios, 31 industries, and 31 languages. It is used for writing outreach emails, replying to inquiry emails, follow-up emails, as well as client background checks, client acquisition analysis, and GEO diagnosis. Multilingual output is supported, with copy organized according to the FABE marketing framework and taking GDPR compliance into account. This article describes its positioning, verified capabilities, activation methods, and typical use cases, and points out limitations in the provided materials, including the absence of installation commands, itemized scenario lists, and platform-specific differences.

Read More
The Ultimate Companion for AutoCAD Electrical: Complete Your Symbol Libraries, Wire Terminations, and Templates in One Go

Community-assisted skill for AutoCAD Electrical engineers, coordinate @user_51a923b7/autocad-electrical, version 1.0.10, category professional, source community. Targeting high-frequency pain points such as missing symbol library/title blocks and messy/mismatched wire and terminal configurations, it provides actionable configuration generation, template production, and troubleshooting checklists, with the goal of standardized, unified, ready-to-use outputs. It can be enabled without an API Key. The article describes its positioning, three core capabilities, applicable scenarios, and notes/caveats, and points out that the official page was inaccessible this time and that the installation command and client differences could not be verified.

Read More
SWOT Strategic Analysis Skills: Turning a Strategic Analysis into a Deliverable Document

The SWOT Strategic Analysis Skill (@engicool/swot-analysis-skill, Version 1.0.0, Publisher: 工程客, Category: business-ops) enforces guided interactions to first communicate fully with users, then completes a four-quadrant analysis of strengths, weaknesses, opportunities, and threats and the deduction of four strategic combinations: SO, WO, ST, and WT. It outputs structured Markdown and web versions for decision-making and demonstrations. This article introduces its core functionalities, verifiable metadata, the typical workflow from its official description, as well as applicable scenarios and limitations. As the official page could not be scraped, and installation commands and configuration fields remain unverified, these aspects are not elaborated here.

Read More