Canvas Visual Design: Generating Posters and Single-Page Visuals with Design Philosophy
Introducing the SkillHub “Canvas Visual Design” Skill (handle: @kunlungrowth/kg-canvas-design, version 1.0.0, source: enterprise, category: design-media, publisher: Kunlun Growth). It uses a design philosophy to drive the creation of posters and single-page visual art, producing original PNG or PDF outputs. The page tag shows requires_api_key=false. The article describes its core capabilities and the official URL, and notes that in verifiable materials there is no SKILL.md body text, installation commands, or differences for Cursor/Codex CLI/Claude Code; therefore no commands are fabricated. It also provides a general enable-and-use workflow as an inference, and highlights limitations and items that require hands-on testing.
Read MoreLao Mo, Master of Self-Evolution Systems: Equipping Agents with a Growable Memory and Skill Framework
"Lao Mo Self-Evolution System Master" is an Agent Skill based on the universal SKILL.md format (version 1.0.0, categorized as ai-agent, published by @user_f67fc7cb, and requiring no external API keys). It targets developers aiming to build self-growing Agent systems or digital partners, decomposing personality, memory, skills, knowledge, and workflows into a five-layer interconnected architecture, and is equipped with three implementation mechanisms: task coverage, behavioral rules, and scheduled specializations. The public version features a built-in standard schema for experience entries (including the evidence_count field and a forgetting decay mechanism), a deduplication and consolidation decision tree, strict security rules (dangerous rewrite interception, path leak filtering, constraint contract state machine, two-question exit gate, and exception degradation), quantitative self-assessment for evolutionary health (the Golden Triangle, A/B comparison, regression testing), scheduling registration guidelines, proactive evolution inspections, out-of-the-box scaffolding (init_memory.py, gen_automation_spec.py, config.example.cfg), a test case suite in evals/evals.json, and an external skill learning analysis method. The official statement declares that it is not bound to any proprietary platform and can be reused across industries. Specific installation commands and platform differences were not verified in this scrape; please refer to the official page for details.
Read Moredsh-desktop-alerts: Windows System Notifications upon Approval and Task Completion
DSH approval cards are only shown inside the window and are easily missed when you switch away or enable Do Not Disturb. dsh-desktop-alerts converts approvals and task completions into native Windows toast notifications. On Windows 11 it uses the "urgent" priority; on Windows 10 it uses the "alarm" category to bypass Do Not Disturb and fullscreen, without interrupting when the app is in the foreground. Clicking the notification opens the dshalerts protocol to return to DSH. Supports notification sounds, repeat reminders, and completion notifications. Settings are under "Desktop notifications"; configuration is in ~/.dsh/dsh-desktop-alerts.json. Requires Windows 10 1809+ and DSH Desktop 0.2.1+. MIT-licensed open-source project.
Read Moredsh-auto-rerun: Automatic Retry and Resume Within Sessions on Model Connection Loss
Maintained by good-saltfish, the DSH host plugin dsh-auto-rerun (v1.8.5) takes over retries upon disconnection-related failures (network issues, rate limiting, 5xx errors, timeouts, etc.). It features Layer A for long backoff and retryUntilSuccess, Layer B for optional follow-up continuation, and Layer C for model-switching retries (disabled by default). It complements the built-in dsh-llm-retry. Install via `plugin_manager install_bundle`. Includes a right-sidebar status and configuration panel. MIT License.
Read Moredsh-voice-dialog: DSH Continuous Speech Dialogue and Automatic Reading
dsh-voice-dialog is a DSH client plugin maintained by ggfgfgf-on (version 0.8.1, MIT license). In the Composer, a single microphone enables continuous dictation and read-while-listening via Edge TTS, with default queue-based delivery and ducking. Parallel organizer, voiceprint CAM++, and MossFormer2 dual-speaker separation are optional. Installation requires DSH 0.2.1-rc.2+, Node.js 22+, and an STT host, with the built-in experimental voice-input bundle disabled. This article, based on the GitHub README, covers the installation commands, usage, and the locations where configuration is stored.
Read MoreFootball Information: Football Hot News Summary Filtered by League and Team Skill
"Football News Digest" (@user_2cc9f73a/football-news-digest) is a community-sourced Skill on SkillHub, categorized under knowledge-management. Version 1.0.0, published by user_2cc9f73a, it carries the tag requires_api_key=false. It scrapes real-time football trending topics via search engines and compiles them into a 5-10 item briefing. Each entry includes a title, a one-sentence summary, and a source link, with support for filtering by league, tournament, and team. Trigger keywords cover Chinese and English queries such as football news, football hotspots, football news (English), Premier League, La Liga, UEFA Champions League, Serie A, Bundesliga, Ligue 1, Chinese Super League, World Cup, and Euro news. The article outlines its positioning, core features, coordinate-based installation method, typical prompt examples, and usage limitations including content timeliness and the upper limit on the number of items.
Read MoreDocument Inspection AI: Read-only Review and Erratum Report for Multi-format Documents
Document Proofreading AI is a community-published Agent Skill: Coordinate: @user_909957f0/doc-proofread; Version: 1.0.1; Category: office-efficiency; requires_api_key=false. Designed for scenarios such as client materials, solution documents, and press releases, it supports read-only proofreading for PDF, Word (.docx), and plain text (.txt/.md) files. It conducts a four-tier progressive review covering text, punctuation, data logic, and style compliance, and offers three optional modes: client data verification, internal solution review, and press release proofreading. It is strictly read-only: it does not modify source files, outputs an independent error correction comparison report, requires confirmation for any modifications, and performs self-verification after changes are made. The article outlines its positioning, core features, installation and activation notes, and typical trigger phrases, and includes the official URL.
Read Moredsh-im-phone-chat: Read-only mirroring of mobile IM sessions on DSH Desktop
dsh-im-phone-chat is a DSH client plugin maintained by FST-EchoFly that works with @xmanrui/dsh-im to display mobile IM conversations from QQ, WeChat, Feishu, and other platforms as read-only bubbles in the left sidebar and main panel. It supports streaming responses, Markdown, LaTeX, and image preview. Recommended installation: `dsh plugin --profile <profile> add dsh-im-phone-chat`; a page refresh is required after installation. Generally, no configuration is needed; for read-only sending, you need to navigate to the conversation or reply via your mobile device. MIT licensed, npm version 1.4.0.
Read Moredsh-rice-mode: The "Unlimited White Rice Mode" Three-Level Switch in DSH
frisk325's DSH workflow plugin dsh-rice-mode uses a three-level capsule in the bottom-right corner (off, this round, always on) to control whether to inject the fixed text "别省 token" into the systemPrompt. It supports local HTTP switching; installation requires running npm install in profiles/web and writing to dsh.profile.bundles, followed by a restart. The switching and injection are cross-platform; on Android, Shizuku can be optionally used for long-press restart. Based on README 1.1.0, DSH 0.2.0-rc.2, MIT.
Read MoreCritical Reading of Medical Literature: Conducting Peer-Review-Level Critical Analysis of a Single Medical Article and Generating a Structured Excerpt Report
A Skill designed for intensive reading of individual medical literature, conducting deep analysis based on peer review standards and generating structured excerpt reports. It covers clinical studies, animal experiments, cell experiments, molecular mechanisms, and mixed-type research, with keywords including GRADE, SYRCLE, and ARRIVE. Released by community publisher user_767538ff, this is an entry for the World Pharmacists Day AI Pharmacy Skill Competition (Version 1.0.0). Sourced from the community and categorized under knowledge-management, it requires no API key. Since official installation commands and usage examples are not provided in the materials, this is stated truthfully in the text, and users are advised to refer to the installation instructions on the SkillHub page.
Read Moredsh-local-search: Enabling Models to Query Local Machine Index Instead of Guessing Paths
This document introduces the DSH plugin **dsh-local-search**, maintained by fishdivinity. On Windows, it searches full-disk file names via **Everything** (es.exe), and searches the content of PDF/Office/EPUB files via **Anytxt**. It exposes five read-only tools to the model, including **local_search_files**. The document covers the installation profile command, dependencies and GUI requirements, configuration entry points, and applicable scenarios. It also notes that the plugin runs with the permissions of the DSH process, and that users should review the source code and licenses themselves.
Read Moredsh-strata-sync: Aligning DSH Models and Context Compression Budgets to Strata Server
dsh-strata-sync is a DSH client plugin maintained by ethan741205. In the settings page, it read-only fetches the current model and verified capabilities from the Strata Server, and after previewing the differences, it backs up and syncs DSH custom models, output limits, and the Agent preset's compaction budget. Optional features include auto-sync on startup and a 60-second runtime mismatch alert. It supports editing summary length descriptions and Dynamic summary target. The plugin is read-only with respect to Strata and does not modify ongoing sessions. Installation can be done via the plugin manager by entering dsh-strata-sync, or via CLI: dsh plugin --profile desktop add dsh-strata-sync --registry=https://registry.npmjs.org/. Before use, you need to configure Strata's OpenAI-compatible provider in DSH Models.
Read MoreArchitectural Deepening: Codebase Scan and HTML Report on Areas for Further Refinement
Architecture Deepening (v1.0.0, by kunlungrowth, category: dev-programming) scans existing codebases to identify opportunities for "deepening," generates a visual HTML report, and then interrogates selected items one by one. It requires no API Key. This article verifies the capability boundaries defined in official materials: it performs only three steps—scanning, reporting, and follow-up inquiry; it does not automatically modify code, nor does it provide public installation commands or reproducible examples. Therefore, no speculation is made regarding directory differences or command-line syntax for Cursor or Claude Code. At the time of retrieval, the download count was 1,216 with 0 installs; given the small sample size, the score of 100,000 has limited reference value. It is suitable for projects requiring review or handover, or those seeking to make implicit architectural constraints explicit, but is not suitable for empty repositories or scenarios requiring function-level call chain analysis.
Read MoreAmazon Industry Benchmarking: Category Star Rating & Negative Review Rate Ranking — Analyzing ASINs within the Category Context
This article introduces the community Skill "Amazon Industry Benchmark · Category Star Rating & Negative Review Rate Ranking" (*@user_507a5056/industry-benchmark*, v1.5.2, *business-ops*). The free tier displays the product's relative positioning regarding star ratings and negative review rates within its category. The paid tier provides category rankings based on three metrics: recent 30-day popularity, negative review rate, and average star rating, facilitating product selection and competitive positioning. The same package offers ARI review analysis capabilities, including ASIN subscription and data collection, review filtering, star rating and keyword trend radars, VOC and deep insight reports, product operation workflows, watch list change monitoring, negative review alerts and workbench, and CSV/Markdown/HTML exports. An ARI API key with the *ari_live_* prefix is required. As the retrieved documentation does not provide installation commands or details regarding multi-environment differences, this has been honestly noted in the article without fabrication.
Read MoreFinance Elite: Structured, Compliant, and Traceable Financial Automation Skill
The Finance Elite (User user_7b2a66ab/finance-assistant-skill) is published by SkillHub user user_7b2a66ab, version 1.0.0, source: community, category: professional, tag: requires_api_key=false. It provides structured, compliant, and traceable financial automation support when users need to handle finance-related tasks, suitable for scenarios including invoice verification, three-way matching, bank reconciliation, aging analysis, cash flow forecasting, tax validation, and financial report generation. The trigger conditions are when user intent involves amount calculation, bill processing, accounts receivable and payable management, tax compliance, or financial data analysis. It is not applicable to general non-financial data processing, unstructured casual conversation, or legal consultations beyond the scope of financial expertise. This article describes the skill's positioning, core functions, enablement method, and applicable boundaries; the materials do not provide official installation commands, reproducible examples, or dependency descriptions, and specifics should refer to the official SkillHub page.
Read Moredsh-wechat-wake: Scheduled Wakeup in WeChat Social Mode
dsh-wechat-wake provides three sets of timed tools—remind_later, remind_list, and remind_cancel—for DSH WeChat social mode sessions, bypassing the restrictions imposed by schedule and bridge in social mode. At the scheduled time, content is delivered to the current chat via follow-up messages or direct messages. Authorization remains handled by WeChatAgent through per-chat selection. This article, based on the repository's README, introduces installation commands, self-check mechanisms, tool behaviors, dependency channels, and security boundaries, making it suitable for WeChat agent scenarios involving companion chat and personality-driven interactions.
Read Moredsh-connect-workbuddy-models: Consolidate local WorkBuddy multi-accounts into the DSH model pool
The DSH plugin dsh-connect-workbuddy-models (v0.1.1, MIT) by DragonHorizon-xtihc aggregates local WorkBuddy login snapshots into two model pool providers—domestic and international. It automatically discovers accounts, reads CycleCapacityRemainPrecise credits and expiration dates, provides expiration reminders and check-in functionality, supports five account rotation strategies including expiry-based switching, and allows account switching and total credit viewing directly below the input box. It can coexist with the single-account dsh-connect-workbuddy. Installation involves adding the profile package.json dependency and bundles configuration, then running pnpm install and restarting DSH.
Read Moredsh-file-integrity-check: Cleanup of DSH workspace sandbox ACL residuals
DSH leaves Low-integrity labels and S-1-4 sandbox ACEs in disk ACLs for Windows Restricted Workspaces, which may persist even after session or workspace removal. The plugin dsh-file-integrity-check (Dofang-C, v1.2.0, MIT) reads the DSH live workspace list, performs read-only scanning of three types of sandbox remnants, repairs and verifies after confirmation, and retains a replaceable rollback record per workspace; supports sidebar and PowerShell CLI, requires Windows 10/11 and DSH≥0.2.1.
Read Moredsh-live-vibecoding: Broadcasting DeepSeek Harness Agents on Bilibili
dsh-live-vibecoding (Deiael, MIT, 0.1.0) is a DeepSeek Harness community plugin that enables Agents to read danmaku, thank gifts, remember regular viewers, and manage song request queues on Bilibili, with voice output via Doubao TTS and streaming through OBS. It includes a live streaming console, a customizable stage with interchangeable skins, and automatic OBS/Bilibili configuration. Installation: dsh plugin --profile desktop add dsh-live-vibecoding, and add it to dsh.profile.bundles. Requires OBS 28+, a Bilibili live room, and a seed-tts-2.0 Key for TTS. Based on the dsh-livestream architecture.
Read MoreTicket Parsing, Data Cleaning, Field Extraction: Transforming Firefox Ticket Text into Structured Data
This article introduces the community Skill "Ticket Parsing, Data Cleaning, Field Extraction" (@user_a10ab3ad/foxtracs, version 1.0.2, categorized under data-analysis) on SkillHub. It reads Firefox ticket text, extracts key fields, and outputs structured data to facilitate subsequent statistics and analysis. The tag `requires_api_key=false` indicates that no API Key is required. This article outlines its core capabilities, applicable scenarios, and limitations. It also truthfully states that the official page could not be accessed during the writing of this article, and installation commands and examples were not included in the available materials. Therefore, no unverified installation steps are provided; users are advised to refer to the official page for details.
Read Moredsh-power-menu-mac: One-Click Refresh, Restart, and Close for the macOS Desktop DSH Title Bar
dsh-power-menu-mac is an MIT-licensed macOS port of the Windows plug-in dsh-power-menu, maintained by DDMUC. It places a power button in the top-right corner of the DSH Desktop session title bar, providing actions to refresh the page, restart the app (with window-ready validation), schedule shutdown/restart, and fully quit the application. Before performing a refresh, it checks whether the plug-in bundle has been modified to avoid revision conflicts. After installation, users must run enable.sh to write entries into dsh.profile.bundles and then restart DSH. The tool is strictly local and Darwin-only, but it can also be used alongside dsh web. Configuration includes a default countdown time in minutes and optional system prompt injection.
Read MoreMini Program Canvas Preview & Screenshot Visual Verification (8-Point Checklist): Regression-Ready Screenshot Checking for Canvas Rendering
A Skill designed for local preview and screenshot visual verification of WeChat Mini Program canvas projects, with coordinates @user_3ff3c268/miniprogram-canvas-verify. Maintained by community publisher user_3ff3c268, it is categorized under dev-programming (Version 1.2.0, Source: community, Tag: requires_api_key=false). It chains a fixed workflow of "starting a static preview service — executing JS screenshots via Selenium — cropping and regression comparison based on scene YAML," and provides an 8-item verification checklist for preview.html (blank screen, lines, centering, instant redraw, export, color values, console, and consistency). Defaulting to port :9099, it supports specifying root directories and ports. Utilizing a pure frontend + Python toolchain with zero business dependencies, it requires no API Key. Trigger words include preview canvas, screenshot verification, visual regression, verify rendering effects, miniprogram canvas verification, and start preview service. This article only documents verified capabilities; since the official page was inaccessible during this session, installation commands, scene YAML fields, and examples are not fabricated. Furthermore, it clarifies limitations, such as not replacing business logic testing and being susceptible to viewport and anti-aliasing effects.
Read Morevideo-downloader-by-browser: Chunked Video Download Skill Involving a Headed Browser
video-downloader-by-browser is a community Skill, located at @user_55bc9e8d/video-downloader-by-browser, categorized under design-media, version 1.0.3. It utilizes a headed browser to open video pages, allowing the user to complete logins, SMS verification, slider captchas, viewing password entries, or definition switching. It then intercepts the actual stream URLs, downloads TS streams / m3u8 / large MP4 segments in parallel, losslessly merges them into a complete MP4, and cleans up the segments after acceptance verification. It is suitable for restricted segmented video sites such as Youku, Tencent Video, iQIYI, and Bilibili. Scraped data indicates a score of 100000, 1217 downloads, and 0 installations. As the official page scraping failed this time, installation commands and client differences remain unverified; therefore, no commands have been fabricated in the text.
Read MoreDSH Plugin Store: Browsing Community Plugins in Harness
The Harness Plugin Store (dsh-plugin-store 0.3.5), a DeepSeek Harness client plugin developed by Chiyang001, offers community plugin discovery, nine filter categories with search, favorites, npm installation mirrors, and format:1 subscription feeds in the sidebar. It also manages plugin installation and enablement/disabling by reading the host manifest. The plugin includes 860 built-in community directory snapshots and offline daily poetry, and supports the mobile bottom bar and a warm theme. Web installation: `dsh plugin --profile web add github:Chiyang001/DSH_Plugin_Store`. Compatible with `>=0.2.1-rc.5 <0.3.0`. This is a community project, not an official store.
Read MoreThe Complete Workflow for Traffic Accident Disputes (by Lawyer Zhang Chenlu): Standardizing Case Handling Procedures into a Skill
An Agent Skill designed for legal practice, released by community developer user_281f816d on SkillHub (Version 1.0.1, Category: Professional). It covers liability determination and review, special types of accidents, disability assessment, compensation schedules, insurance negotiation, dual compensation for work-related injuries, litigation and mediation, and includes guidelines for on-site evidence collection by parties. This article introduces its positioning, scope of capabilities, installation and activation methods, typical usage scenarios, and applicable boundaries.
Read More