dsh-theme-strinova: A Strinova Theme for DeepSeek Harness
dsh-theme-strinova is a fun theme package for DeepSeek Harness, released by its maintainer nizihan29 (v0.1.0, MIT License). It features light/dark themes with a warm amber-gold color scheme, randomized character-card rotations (including lines and chibi-style facial expressions), floating particle effects, and a customizable sidebar logo. Asset credits indicate the materials originate from the Strix Wiki. To install, select the official npm registry within the DSH plugin and enter dsh-theme-strinova, or search for the package with the same name in the Plugin Marketplace; no additional CLI build commands are required beyond what’s provided in the documentation. Designed for developers who want a Strix-style aesthetic within Harness over the long term—review the source code and license notice before installing.
Read Moreai-rules-sync: batch parsing, synchronization, and management of AI rules configuration
ai-rules-sync (Coordinates: @org-k6bcf2fv/ai-rules-sync, Version 1.0.0, Source: enterprise, Category: ai-agent, Published by: Skill训练营) is a management Skill designed for AI coding tool configurations. It is used to parse, sync, and manage AI rules, skills, commands, and sub-agent configurations, supporting batch processing and format standardization. It addresses issues such as scattered configuration sources, inconsistent formats, and reliance on manual copying for synchronization by integrating rules, skills, commands, and sub-agent configurations into a single parsing and synchronization workflow. No API key is required for enablement (Tag: requires_api_key=false). The SkillHub page currently displays a rating of 100000, 92 downloads, and 0 installations; as there is a discrepancy between this and earlier leads regarding ratings and download counts, the page data shall prevail. The documentation does not provide independent installation commands, configuration fields, or reproducible examples. Typically, it is enabled by placing files in the corresponding tool's Skills directory following the general SKILL.md format, though the official page should be consulted for specifics, and direct replication is not recommended. Applicable scenarios include aligning scattered configuration sources regularly, standardizing inconsistent formats, and updating configurations in batches. Limitations include sparse documentation, version 1.0.0 only, and zero installations; it is recommended to validate synchronization results in a small scope first. Additionally, for Skills marked as "enterprise," it is necessary to confirm that configuration content is safe to leave the current environment before use.
Read MoreVideo Deconstruction: Analysis and Imitation Skill for Single Short Videos
This article introduces the "Video Breakdown" Skill developed by the SkillHub community, which can deconstruct a single video to generate imitative copywriting. It verifies basic information such as its coordinates, version, category, rating, and download count, and notes that an API key is required. As specific installation commands and invocation details could not be scraped from the official page, users are advised to refer to the official page for accurate information.
Read Moredsh-plugin-preflight: Nine Preflight Gates for DSH Plugin Installation
dsh-plugin-preflight is an admin-security plugin for DSH maintained by NingZhenPeng. It runs nine pre-checks (with thresholds defined by policy.json) before executing pluginManager.installBundle; if any check is rejected, the installation is aborted and an audit log is written. The plugin supports hot-switching between enforce/warn modes, default blocking on pre-check failures, and custom rules. Based on the GitHub README and package.json version 0.15.4, this article provides official installation commands for desktop and web, as well as typical configurations and an explanation of known bypass paths.
Read Moredsh-personal-system-rules: Integrating Personal Rules into System Prompts with Three Agent Guards
dsh-personal-system-rules is a DSH workflow plugin maintained by NingYuanLee (v0.6.0, MIT). It registers PERSONAL-SYSTEM-RULES.md as the system-prompt segment user:agents-md and can optionally enable silent step reminders, loop detection for identical tool calls (remind/stop), and a stale-todo guard. Rules and volatile guard parameters can be hot-updated under Settings → Personalization Rules. Installation is supported via npm, git tag, or the Release tarball. Requires Node >= 22.19.0 and dsh 0.2.1-rc.2 up to <0.3; it has no effect under the minimal preset. Installing plugins runs third-party code with user privileges; review the source before use.
Read MoreSCI Literature Brief (AI Practical Camp): Automating PubMed and CNKI Searches into a Weekly Literature Bulletin
The "SCI Literature Briefing (AI Practical Camp)" is a literature hotspot briefing Skill designed for critical care medicine education scenarios. It was published by user_4f291e59 on SkillHub, with version 1.0.0, categorized under education, and marked as requires_api_key=false. It provides four capabilities: searching English literature in PubMed and outputting bilingual Chinese-English titles and abstracts, searching Chinese literature in CNKI, generating briefings saved as local Markdown files, and automatically writing to IMA Notes. It also supports automatic weekly execution through WorkBuddy scheduled tasks. The official documentation provides six trigger words: literature briefing, hotspot search, weekly report, weekly briefing, recommended literature, and critical care education literature. The article also explains installation and activation information, typical usage patterns, applicable scenarios, and important considerations.
Read MoreRuanbaobao Software Copyright Application Tool: Generate Software Copyright Registration Documents with rzcode-mcp
The "Ruan Baobao Software Copyright Application Tool" is a community-sourced Skill located at @user_68cbafa7/rzcode-workflow-skill (version 0.8.0, categorized under office-efficiency). It depends on rzcode-mcp to generate three documents required for local software copyright registration: software user manual, source code documentation, and registration application help form. The tag "requires_api_key=true" indicates that an API key must be configured before activation. Starting from the pain points developers face in organizing registration materials, this article introduces the Skill's positioning, verified core functions, coordinates, and version information. It candidly notes the absence of official installation commands, configuration fields, and reproducible examples in the scraped data; therefore, it only describes the usage logic without fabricating commands. It also points out its limitations: it relies on MCP availability, the generated results still require manual verification of page numbers and screenshots, the community installation count is 0, and there are conflicts regarding ratings and download counts between task clues and scraped metadata. Finally, the official URL is provided.
Read Moredsh-jev-mcp: Integrating Jev Judgment Tools into DeepSeek Harness
dsh-jev-mcp is a DSH community plugin maintained by NingYuanLee. It launches @jkudish/jev-mcp within the Harness, registering judgment tools such as verify, review, and gate as mcp__jev__*, with the API Key stored in the DSH credential store. The article describes the installation command, the jev skill, and the configuration fields apiKeyEnv/serverName/model/baseURL, along with behaviors such as not registering tools when credentials are empty and a 2-minute invocation timeout. It also lists the dsh and Node version requirements, as well as common troubleshooting steps.
Read Moredsh-subagent-guard: Restricting DSH Subagents with Sandbox Hard Limits, Command Blacklists, and Delegation Toolsets
DSH sub-agents inherit the parent session's permission snapshot upon creation, with approvals permanently set to 'never'; if the parent session possesses 'danger-full-access', the sub-agent is likewise unsandboxed. NightPainters' dsh-subagent-guard provides three layers of safeguards over this snapshot: a hard cap on the agent/created sandbox (subagentSandboxCap), a ctx.tools.guard denylist for destructive commands (including script inspection), and a delegate tool that fixes the persona and tools while binding policies in a fail-closed manner. The installation command is `dsh plugin add dsh-subagent-guard`. Please note that prior to version 0.3.7, there is a risk of the third safeguard being canceled, regular expressions are not sandboxed, and the plugin runs with the permissions of the DSH process.
Read MoreDSH-Handoff: Writing Handoff.md in Extended Sessions, Continuing from Read-Only Docs in New Session.
dsh-handitoff is a DSH workflow plugin maintained by NFdrybattery. In the previous session, the model is instructed via /handitoff to generate a HANDOFF.md containing 0–12 sections, with Section 12 appended verbatim by the plugin from the session log. The new session can then take over using /handitoff resume or a read-only "Handoff & Continue" document. Documents are stored in the workspace at .dsh/dsh-handitoff/, with configurable options such as retentionDays, recentRounds, and language. Requires DSH 0.2.1-rc.x and Node ^22.19.0 or >=24. Installation is primarily via Settings → Plugins or link: local paths, under the MIT license.
Read MoreHistorical Figures' Legendary Life Narrative Generator: Transforming Historical Records into Structured Long-Form Narrative Articles
' The Legendary Life Narrative Generator for Historical Figures (@user_31834641/historical-figure-narrative) was released by community developer user_31834641. Version 1.0.0 falls under the content-creation category and is tagged requires_api_key=false, so no additional API key configuration is needed. Targeting WeChat Official Account writers and short-video oral narration creators, it breaks historical figure writing into four steps—historical source research, narrative-angle design, chapter structure, and drafting plus polishing—to produce narrative articles with a strong sense of story and recognizable spiritual symbols, replacing the "quotable lines + reading reflections" template. The article outlines the verified feature scope, supported integrations and how to enable the tool, as well as typical usage patterns, and notes cautions including that it is community-sourced with an install count of 0, that key historical facts need manual verification, and that its historical source retrieval capabilities are unspecified. During this writing session, both official links were unreachable; scores and download counts are based on retrieved data (score 100,000, downloads 901).
Read MoreComprehensive Guide: Configuring a Local Python Quantitative Computing Skill for AI
QuantAll (Full A-Share Analysis) is a community Skill published on SkillHub (Coordinate: @user_7641c221/quantall, Version 1.0.4, Category: professional). Positioned as a vectorized computation engine for the entire stock market, it provides AI clients with a local Python computing environment via MCP. By enabling AI to write vectorized code, it completes factor calculations, strategy backtesting, IC analysis, and GPU visualization for over 5,000 stocks within seconds, transforming AI from an information retriever into a data computation engine. It is triggered only when users mention quantitative keywords such as backtesting, factor analysis, IC analysis, or stock selection strategies. Enabling it requires creating a Python virtual environment (300MB+) on the user's computer, installing the quantall library, modifying MCP configurations, starting a local HTTP service on localhost:8686, and creating configuration and launch scripts; all local operations must obtain prior user consent. UpdateStock is an optional auxiliary MCP that requires a tushare API and is not essential. The page displays a score of 100,000, 849 downloads, 0 installations, and the label requires_api_key=false. It is suitable for users with clear quantitative analysis needs who are willing to maintain a local environment.
Read Moredsh-taste-loop: DSH Taste Closed-Loop and Versioned Rule Consensus
dsh-taste-loop is a DSH host workflow plugin maintained by NelsonLongxiang (version 0.2.1, MIT license) that provides tools such as taste_rules/case/record/queue/update. It supports version-controlled rule self-service proposals, activation via endorsement from two independent sessions, entry into needs_arbitration upon double dissent, and manual resolution via arbiterSessionIds. Configuration requires verifierRepo, inboxHome, the host/port of taste-mcp, and marketGatewayKey. Install via `dsh plugin --profile web add` with a local path. It is suitable for multi-session agent scenarios requiring auditable taste protocols. The plugin runs with DSH process privileges; review the source code before installation.
Read Moredsh-plugin-agnes-image: Integrating Agnes Text-to-Image, Image-to-Image, and Text-to-Video for DSH Agent
The neil-ji maintained community DSH desktop plugin **dsh-plugin-agnes-image** (v0.3.1), built on Agnes Image / Video 2.5 Flash, provides Agents with four tools: `agnes_text2img`, `agnes_img2img`, `agnes_video`, and `agnes_video_status`, with results saved to the working directory. The API Key is written to the DSH credential store via **Settings → Agnes image generation**. To install, paste the GitHub repository address on the plugin page. This is an unofficial wrapper and requires a DSH desktop client with tools, `read_image`, and credential services.
Read MoreJiadou AI: Marketing Agent Skill for Social Media Operations
Jiadou AI (@org-ri083vxl/jiadou, version 1.0.16, released by WeBank Internet, categorized as content-creation) is an enterprise-level marketing intelligence agent platform designed for AIGC content generation and social media operation scenarios. Verified capabilities include video generation, AI commercial photography, digital human broadcasting, product scene images, product posters, Xiaohongshu (Little Red Book) note generation, and social media publishing on platforms such as Douyin, WeChat, and Xiaohongshu. The tag requires_api_key=true indicates that an API Key must be configured before invocation. The official description does not provide specific installation commands or configuration fields; this article describes the activation method following SkillHub's general process and notes differences in support across various Agent environments. The SkillHub page shows a rating of 100,000 and 887 downloads. It is suitable for developers who need to integrate marketing material generation and multi-platform distribution in Agent workflows; it is recommended to confirm authentication, quotas, and other details before integration.
Read MoreWord Document Polishing Tool: Input docx, expand and polish based on existing text
This Skill is designed for Word document refinement. Users provide a local .docx file path; the Skill reads the document's content and rewrites it based on the existing text, rather than simply copying and pasting the entire document. The confirmed features include reading docx files, expanding existing text, and it does not require an API key (requires_api_key=false). No information was found regarding the retention of images, tables, and formatting, nor did it mention generating image descriptions. In terms of installation, the documentation did not provide commands or configuration fields, but confirmed the coordinate @user_5162597a/12345678, category office-efficiency, source community, and version 1.0.0. The data snapshot shows 100000 ratings and 876 downloads but 0 installations, indicating that the number of actual users may be very low. The goal is to polish and enhance existing text for clarity and impact.
Read Moredsh-plugin-notify: Agent Pops Native Windows Desktop Cards While Waiting for Your Reply
dsh-plugin-notify is a DSH client plugin maintained by nebula9i (version 0.1.0, Windows only). It pops up native desktop card notifications when an Agent is waiting for you via ask_user_question, plan reviews, or permission approvals. The cards feature no response buttons; clicking one returns you to the chat to answer. It supports multi-Tab, multi-chat stacking, and countdowns within the same session. Installed via package.json link dependencies and pnpm install within the desktop profile, it allows configuration of countdownMs, sound, maxStack, and more via cordis.patch.yml. The plugin attaches to the Cordis waterfall as an observer without altering chat behavior.
Read MoreWorkBuddy Workspace Migration & Recovery: Rebuilding Workspace from Sync Drive After Switching Computers
The WorkBuddy environment migration and recovery is a community-released Agent Skill (coordinate: @user_a9bed549/workbuddy-env-migrate, version 1.0.0, category it-ops-security), created by the community for users to restore a complete WorkBuddy environment from sync drives (Jianguoyun/坚果云, OneDrive, Dropbox, etc.) after switching computers. It decomposes the migration into a three-step workflow of backup → package → release, covers three types of assets: skills, configurations, and memory, and provides diff reports and merge-release, as well as scanning diagnostics and repairs. The article introduces the Skill's positioning, core capabilities, trigger words, and typical recovery workflow, and outlines applicable scenarios and limitations: being community-sourced and requiring user self-review; the restoration hinging on the completeness of sync-drive backups; and metrics such as ratings and downloads conflicting across materials and should be verified against the official website.
Read MoreOTC-GEO Campaign-02-GEO Q&A Content Generator: Generating AI-Citeable Q&A Content for Real User Queries
A GEO content skill for models including DeepSeek, ERNIE Bot, Tongyi Qianwen, and Kimi, produced by Baitong AI Service Platform, classified under content-creation, version 1.1.0. It takes a real user question as input and outputs an answer-first 40-60-character answer block with evidence traceability, E-E-A-T annotations, and extractable JSON-LD, intended for filling gaps in FAQ-structured content. Trigger words include GEO Q&A, Q&A content, AI Q&A, structured Q&A, Q&A generation, content gap-filling, answer-first, and FAQ generation. No API key is required. The article explains its positioning and four verified capabilities, notes that installation commands and official examples are absent from the crawled materials, provides the activation sequence and typical usage, and cautions that model inclusion is not guaranteed, annotations do not constitute qualification endorsements, SkillHub shows 839 downloads with 0 installs, and community verification remains insufficient.
Read Moredsh-plugin-superagent: DSH Cross-Project Task Console
The DSH workflow plugin maintained by NattoCB provides a project matrix on local port 30808, offering persistent tickets, explicit project routing, replaceable DSH session execution, and manual acceptance. It requires Node.js 22+ and DSH 0.2.0-rc.2, installing the bundle via the Plugin Manager by submitting a GitHub repository URL. It does not include LLM auto-routing or a Watchdog.
Read Moredsh-sidebar-todo: DSH Sidebar Todo and Agent Read/Write List
The DSH client plugin dsh-sidebar-todo, maintained by Natte, provides a "To-Do" entry and main panel in the sidebar, featuring eight checklist groups, recurring tasks, a month-view calendar, categories, a note dock, and filtering. Data is written by default to todos.json under the profile. It registers personal_todo and personal_note, allowing the agent and user to share the same local JSON file. Recommended installation: dsh plugin --profile web add dsh-sidebar-todo; optional staleDays and dataPath settings are available. Host engines.dsh version requirement: >=0.2.1-rc.2 <0.3.0-0. MIT licensed.
Read MoreLaunchCheck v3: Run a system-level quality audit before going live
LaunchCheck v3 is an AI quality-assurance skill for the pre-release phase, released by Shanghai Yingchu Industrial Co., Ltd. It performs system-level pre-launch checks and post-release inspections on live websites, mini-programs, and apps, covering functional acceptance testing, bug and defect detection, privacy policy and data security compliance checks, and performance and accessibility testing. It outputs acceptance reports with graded verdicts and remediation tickets. The product offers three tiers: free health check /lc-free (¥0, health score, risk level, three key issues); standard /lc (¥4.99 per run, a 15-dimension report with work orders); professional /lc-pro (¥14.99 per run, adding GDPR, CCPA, HIPAA, and China’s Multi-Level Protection Scheme (MLPS) compliance libraries). No API key is required, with pay-as-you-go billing. The article introduces its positioning, features, integration methods, and applicable boundaries.
Read Moredsh-fish-reflection: Maintaining an Auditable Reflection and Personality State Layer for the DSH Main Session
dsh-fish-reflection is a plugin created by Najiya5 for DeepSeek Harness (v0.8.12, MIT licensed) that maintains 3–7 self-authored hierarchical principles for selected main sessions, along with external loops for mental state, health, and balance, as well as history growth archives. It supports scheduled or manual reflection, with optional note intake from Bilibili and Douyin. It is not a factual memory plugin. Installation is via local directory bundle: `dsh plugin --profile desktop add ./dsh-fish-reflection`. Before use, please note that the default model and optional external analysis APIs will receive summaries and materials, and sensitive configurations are stored in plaintext at ~/.dsh/fish-reflection/.
Read Moredsh-aichat-bridge: Exposing DSH Sessions via OpenAI-Compatible API for Game AIChat Integration
dsh-aichat-bridge is an internet-enabled DSH tool plugin maintained by Naiqey-Star: it exposes a Harness session with a fixed sessionId on the local machine as an OpenAI Chat Completions endpoint. The Mod's system segment replaces the DSH persona in full; agent turns are executed by DSH, and files within readRoots can be accessed read-only via read_file. It supports a dedicated loopback port 47821, Bearer token authentication, workspace attach, and a /health self-check; it is typically used to integrate with Chill With You AIChat. The install command is `dsh plugin --profile <profile> add dsh-aichat-bridge`, released under the MIT license.
Read MoreGMAI - Company Registration Consultant Master: Introduction to the Equity and Tax Structure Preliminary Screening Skill
This article introduces GMAI — Company Registration Advisor, published by Guomeng Digital Technology on SkillHub, described by its official brief as an initial screening of equity and tax-burden structures. The article verifies the coordinates, version, source, category, publisher, paid label, and page statistics. It notes that the materials do not provide installation commands or reproducible examples, and reminds readers that the initial screening cannot substitute for professional advice.
Read More