dsh-proactive-chat: Leveraging Proactive Conversation Continuation to Make QQ Private Chats More Like Continuous Conversations
dsh-proactive-chat (more human-like chatting) is a DSH plugin maintained by shuaig22. Utilizing dual-rhythm proactive conversation (hot/normal modes), typing delay simulation, cooldown and nighttime do-not-disturb features, along with local personality files and NapCat OneBot, it naturally follows up with brief responses after the owner initiates an active chat. Based on the GitHub README, this article covers the working principle, configuration options, installation commands, and operational notes.
Read MoreAI Stock Picker: High Growth, Low PEG Expectation Gap Radar
The AI Stock Picker (cbxh-stock-picker) is a stock-selection research Skill for the A-share market and extensible markets, published by user_d2258560, sourced from the community, categorized as professional, and currently at version 1.0.3. It integrates early detection of hot trends, upward earnings forecast revisions, PEG valuation constraints, and historical candidate pool tracking into a single framework, producing reviewable and continuously trackable candidate pools rather than an automated trading system. Core capabilities include three-caliber PEG comparison (1Y Growth PEG, 2Y CAGR PEG, and TTM PEG, with 2026E/2027E PE as reference), earnings growth driver decomposition, price-in checks across five windows (1W/1M/2M/3M/6M), candidate pool continuity tracking, and self-contained HTML reports. The typical workflow begins by building three independent candidate pools: a heat ranking list, a theme growth pool, and an earnings upgrade pool. These are then validated across multiple sources using Wind MCP, Eastmoney Miaoxiang, heat rankings, and AlphaPai. Finally, results are written to data/candidate_history.csv and output as outputs/cbxh_stock_radar_YYYYMMDD.html. Full operation relies on external data sources including alphapai-research, Wind Financial Data/Wind MCP, and A-share real-time heat rankings.
Read MoreMonthly Customs (dshjieqi-card): Twenty-Four Solar Terms Dots Below the Title Bar in the Dialog Area
Yueling is a DSH plugin maintained by SDSAEWQ (npm package: `dshjieqi-card`). It displays hoverable, expandable dots under the conversation title bar. Locally calculating the exact transition times of the 24 Solar Terms, it displays the Day Pillar, 72 Pentads, health tips, chronology, and Message Hexagrams. It supports coin-tossing divination with a 7-day history review and includes a built-in `jieqi` skill. It operates completely offline and does not invoke any models. To install, enter `dshjieqi-card` or `github:SDSAEWQ/yueling` on the plugin page, or install via the jsDelivr market index. MIT License; requires Node ^22.19.
Read MoreReview of Kindergarten Class-based Curriculum: Transforming Course Sharing Materials into Usable Teaching Research Feedback
This article introduces the community skill "Kindergarten Class-Based Curriculum Review", which is designed for instructional managers, teaching research group leaders, and teachers. By inputting class-based curriculum sharing materials, it performs diagnostic analysis across six dimensions—curriculum starting points, experiential growth, curriculum context, children's subjectivity, teachers' curriculum awareness, and implementation risks—to generate comments and actionable improvement suggestions that can be directly utilized in school-based teaching research contexts. Furthermore, the article details its positioning, origins, core capabilities, usage methods, and applicable scenarios, while noting the aspects that are absent from or could not be verified in public sources.
Read MoreDSH Temporary Workspace Plugin: One-Click Creation, Sidebar Management, and Secure Cleanup
**Introducing the community client plugin @s0meb0dy3/dsh-scratch-workspace:** Create and manage temporary workspaces in the DeepSeek Harness sidebar, with status display and an idle countdown. On cleanup, it performs validation and moves items to the system trash, retains session archives, and supports optional idle auto-expiration. It includes npm installation steps, typical usage examples, configuration options, and security boundary details. Requires DSH 0.2.1 or higher.
Read Morecompanion-memory: Building Local Personas and Memory for DSH Roleplay Dialogues
companion-memory is a DSH memory plugin maintained by Root-7-KHX (v1.2.1, MIT). It provides independent personality Markdown, affinity, and active memory for each conversation, caches raw messages for approximately 48 hours based on chat time and archives them to a local knowledge base, supporting two-stage retrieval and context compression. New conversations default to disabled and can be enabled with one click from the sidebar. Installation requires obtaining a direct link from GitHub Releases and executing `dsh plugin add`, or installing from a link via the plugin marketplace. Suitable for long-term role-playing; review the source code before use, and data is stored in ~/.dsh/companion-memory.
Read MoreHealthy Shrimp 14-1: Upload Food Images, Identify Food, and Provide Nutritional Components and Dietary Recommendations
When building diet logging or health management applications, the most difficult step is converting "a user taking a photo" into structured nutritional data. Developing a custom recognition and calculation pipeline requires maintaining food databases, configuring prompts, and handling various query formats, resulting in a long implementation cycle. Health Shrimp 14-1 solidifies the pipeline of "upload image → identify food → output nutritional components → provide dietary suggestions" into a Skill that developers can use immediately upon integration. The following section introduces its positioning, capabilities, installation status, and usage considerations. **What is this?** Health Shrimp 14-1 is a food nutrition analysis Skill located at `@user_d93c4aed/sc-food-nutrition-analyzer`. It is Version 1.0.0, sourced from the community, and published
Read Moredsh-mem0-plugin: Integrating mem0 Long-Term Memory into the DSH Process
Rokyaria's dsh-mem0-plugin runs the mem0ai TypeScript SDK directly within DeepSeek Harness, providing eight tools including mem0_add and search. Configuration is stored in cordis.patch.yml, with Elasticsearch 8 serving as the reference vector database deployment and pluggable embedding support. Based on the official README, this article covers the differences from an independent mem0 service, installation commands, minimal configuration, and notes regarding LM Studio. As the plugin runs in the same process as dsh, source code review and adherence to the MIT license are required before installation.
Read Moredsh-quiet-hours: Adding Workday Quiet Hours to WeChat Bots
dsh-quiet-hours is a DSH workflow plugin that intercepts model calls at the agent/pre-step stage within WeChatAgent WeChat sessions. It is designed to send fixed replies via hooks only during configurable time slots on weekdays, allowing users to modify the default text. For private chats, it replies to all messages, while for group chats, it defaults to replying only when mentioned (@). It includes cooling-down periods and hourly rate limits. Outside of the configured time slots, the plugin automatically restores standard AI interactions. This plugin supports installation via the web profile, command line, and the DSH Desktop plugin marketplace. It requires Node ≥20, DSH 0.2.0-rc.2, and WeChatAgent 0.3.9.
Read MoreRespire DSH Plugin: Integrating Respire Memory into Harness
The @rsrsai/dsh from risense-ai is a native memory plugin for DeepSeek Harness. It calls the rsrs CLI through systemPrompt and agent/pre-step, agent/turn-stopping hooks, with the engine handling storage and synchronization. Requires Node.js 24+, verified Harness 0.2.1-rc.5, and rsrs runtime. It is installed via plugin_manager install_bundle, not through direct npm installation in the terminal. The documentation covers migrating old bundles, marketplace index updates, and key validation points for recall/store functionality after installation.
Read MoreDSH-Agent-Team-Scheduler: A Manually Enabled Task Distribution Queue for Agent Teams
dsh-agent-team-scheduler is a DSH workflow plugin (v0.1.2, MIT license) maintained by Rinai-R. In Agent Teams, the Lead registers tasks via team_dispatch_set; after a human runs /dispatch on, ready tasks are assigned to idle teammates; the default WIP limit is 2, and assignments wait if write scopes overlap. Requirements: DSH >=0.2.1-rc.5 and <0.3.0, with Agent Teams enabled in the same profile. Installation: named dsh-agent-team-scheduler or via `dsh plugin --profile web add`. Re-enable after restart; it does not create teammates or worktrees, and scope is only a hint.
Read MoreAI ZhangLe--Financial Analysis and Information Query: Routing Financial Analysis and Information Retrieval to Corresponding Tools
"AI Zhangle - Financial Analysis and Information Query" is a community Skill on SkillHub, located at @user_d9eea1c5/zhangle-financial-analysis, version 1.0.2, categorized as professional with the tag requires_api_key=true. It targets financial analysis and information query scenarios; when users ask questions related to financial market analysis, information retrieval, or analytical diagnostics, it first routes to the corresponding tool based on category before returning professional results. Its core capabilities include analytical diagnostics, market insight analysis, and information retrieval. The article outlines its positioning and capability boundaries, noting that official materials lack verifiable installation commands and configuration fields, and do not detail the integration differences for Cursor, Codex CLI, or Claude Code, requiring confirmation from the publisher. It also cautions that the Skill relies on an API Key, originates from a community account, and based on publicly available platform data (Rating: 100000, Downloads: 1238, Installs: 0), actual adoption is difficult to determine, meaning output conclusions must be independently verified.
Read MorePatent Protection and Intellectual Property Management Skill: Drug Patent Portfolio Layout and FTO Risk Assessment
"Patent Protection and Intellectual Property Management" is a professional classification Skill under the enterprise source on SkillHub, located at @baitongai/patent-protection-and-intellectual-property-management, version 1.0.0, published by Baitong AI Service Platform. The metadata indicates requires_api_key=false, so no API Key configuration is needed before use. This Skill is designed for the product lifecycle management scenarios in the prescription drug treatment areas of oncology, immunology, and rare diseases, covering five directions: comprehensive patent landscape analysis, FTO (Freedom to Operate) risk assessment, patent protection reinforcement, infringement defense and patent continuation strategies, and intellectual property commercialization pathway planning. Trigger scenarios include analyzing patent layouts, assessing infringement risks, developing patent continuation strategies, and planning licensing and collaboration pathways. The article explains its differences from general patent search tools, provides typical usage organized by task type, and highlights three limitations: the official introduction only declares applicability to specific prescription drug areas, patent legal status requires retrieval date annotation, and FTO conclusions do not replace legal opinions from patent attorneys or lawyers. Installation commands and detailed configuration were not provided due to the official page and API address being inaccessible during writing; only verifiable coordinates, version, source, and classification information are given. Readers are advised to obtain installation instructions from the official address before using it in Cursor, Codex CLI, or Claude Code.
Read MoreSciDraw AI Scientific Figure Generation: Transforming Paper Methods into Iterative Research Visuals
SciDraw AI Scientific Illustration Generation Skill introduction. Coordinate @user_95a90f81/scidraw-ai-scientific-illustration, version 1.0.1, source community, category design-media, tags requires_api_key=true. The function is to transform paper methodology, grant technical roadmap, experimental workflows, mechanisms and model architectures into iterable scientific illustrations. The workflow includes official API generation, task querying and result downloading. The API base URL is configurable, defaulting to https://sci-draw.com/api/v1. The documentation does not provide full SKILL.md text and installation commands; it can only be enabled by following the official page instructions, and users must obtain an API key themselves. At the time of retrieval, the page displayed a rating of 100000, 1386 downloads and 0 installations.
Read Moredsh-obsidian-notes: Read/Write Local Markdown Vault from DSH Web Sidebar
dsh-obsidian-notes (revive123456, 0.5.2, MIT) provides a Web right-column Notes panel for DeepSeek Harness 0.2.1+: it points to a local Markdown vault with line-numbered preview and editing, 13 callout types and aliases, tables and wikilinks, screenshot pasting as Obsidian embeds, @ path to reference an entire document or selected paragraph to Ask/Explain in the conversation, and supports session/workspace batch archive and deletion. Installation requires a host restart; use `dsh plugin --profile web add dsh-obsidian-notes` via CLI. No backlinks or graph view.
Read MoreAssistant for Investment Banking Regulatory Responses: Standardizing the Deconstruction and Drafting of A-share Inquiry Replies
A Skill designed for A-share IPO, refinancing, and M&A restructuring project teams, leveraging the IMA Investment Banking Law & Regulations Database and IPO Practice Case Database to assist in decomposing inquiry and internal review feedback questions, constructing response frameworks, drafting suggested response texts, and predicting follow-up questions. This article introduces its positioning, core features, activation methods, and applicable scenarios, while reminding users that it is an auxiliary tool and final responses require review by the project team.
Read MoreLegal Daily: Aggregating same-day Chinese legal news into an HTML dashboard
Law Daily is a community-maintained Agent Skill, version 1.0.0, categorized as professional, and does not require an API key. When its trigger phrases are matched, it uses WebSearch to retrieve publicly available legal information from sources such as the National People's Congress, the State Council, the Supreme People's Court, and XinhuaNet. It generates a responsive single-file HTML dashboard organized into five sections: Legal Updates, Product Releases/Updates, Industry Trends, Research Papers, and Tips & Insights. This article describes its purpose, capabilities, installation and activation methods, typical usage, applicable scenarios, and caveats. Specific installation commands and command-line usage were not provided in the captured materials and should be referred to the official page.</think>
Read Moredsh-bili-danmaku-Lite: A Lightweight Plugin for Bilibili Live Danmaku in DSH
dsh-bili-danmaku-lite from rentaoyan0628-collab is a Bilibili live streaming danmaku (bullet chat) plugin for DeepSeek Harness (v0.5.0). It supports read-only room browsing and danmaku listening, room entry after manual approval, authorized danmaku sending, persistent listening with webhooks, optional AI-powered auto chat companion and speech-to-text transcription, as well as offline highlight timestamp analysis. Room entry and message posting are human-gated by default, with configuration and credentials stored in $DSH_HOME/bili-live. Based on the official README, this article describes installation commands, dependencies, typical usage, and precautions.
Read MorePII Masking Review: Pre-release Compliance Gate for Skills
A pre-release personal information sanitization review skill designed for skill definitions, code repositories, and data directories. Published by community developer user_65c8c185 on SkillHub, version 1.6.0, categorized as it-ops-security, and requires no API key. It employs a two-layer execution mechanism: L1 automated sensitive scanning supports three input modes (directory, single file, and text); L2 six-dimension compliance review generates a prioritized issue list, remediation priorities, and a standardized audit report. Exit codes can be directly integrated into CI pipelines as release gates, and compliance bases support optional real-time verification of original legal text. The tool exclusively performs review, classification, and remediation suggestions to generate reports; it does not execute data sanitization and is not intended for providing legal advice. Manual review is recommended for high-sensitivity scenarios.
Read MoreIntroduction to Trello Kanban Card Management Skill
Based on the standard SKILL.md format, this Skill is located at @clawhub_steipete/trello (Version 1.0.0). Sourced from clawhub and published by steipete, its upstream repository is https://clawhub.ai/steipete/trello. It interacts directly with Trello via the REST API using curl + jq, eliminating the need for a web interface. It covers six core operations: listing all boards, viewing lists and cards, creating cards, moving cards across lists, adding comments, and archiving. This Skill is ideal for scenarios requiring the writing of tasks to Trello boards, batch viewing of all cards within a specific board, or the scheduling of cards between multiple lists. Trigger words include Trello, board management, card creation, Trello API, task boards, curl Trello, and board archiving. The tag requires_api_key=true indicates that a Trello API Key must be configured before use, and usage is subject to Trello API rate limits. Official URL: https://api.skillhub.cn/org-b6kzcl68/trello; SkillHub Page: https://www.skillhub.cn/skills/clawhub_steipete/trello.
Read MoreWhat Sub-Agents Are Doing Now: Returning Progress Written by Sub-Agents to the Parent Agent
The DSH parent agent often fails to see sub-agents spawned by `subagent`/`subagent_fork` via `list_agents`, and their status is typically listed as running/inactive, making it impossible to distinguish between completion, hangs, or failures. The community plugin "what-sub-agents-doing-now" (by REALOIYO, MIT) performs read-only retrieval of step counts, last tools, idle status, and self-written narrations from the sub-agent registry directory, featuring non-shadowing and zero-configuration. It supports installation via desktop GitHub URL paste or `dsh plugin add what-sub-agents-doing-now`. While `spawn_teammate`, `subagent`, and `subagent_fork` have been tested, end-to-end verification for workflow/ralph remains pending.
Read MoreAsideNote: Jot down notes alongside the DSH session
"dsh-aside-note" is a conversational note-taking plugin developed by Rayle8 for DeepSeek Harness. It provides an entry point at the top of the conversation and a right-side panel, supporting workspace/session-level checklists, completion partitioning, automatic draft saving, and one-click forwarding to the input box. Data is stored in the current browser's localStorage and does not sync across devices. Installation uses the official command `npx dsh-aside-note install`, which is idempotent and typically does not require restarting DSH. Released under the MIT license, it is suitable for scenarios where todos need to be temporarily suspended during collaboration without interrupting the conversation.
Read Moredsh-film-analysis: Film Screen Measurement, Eight-Dimensional Analysis, and Automatic Film Pulling in DSH Web
dsh-film-analysis is a workflow plugin released by Raven-Jackdaw for DeepSeek Harness Web: it bundles ffmpeg/ffprobe and offline Chinese OCR, supports visual metrics measurement, eight-dimensional audiovisual language forms, annotation and prompt generation, automatic storyboarding, in-panel AI frame-by-frame analysis and whole-film reports, and can export Markdown and shot-list CSV. It is installed via plugin_manager in a cordis session (npm package name dsh-film-analysis or a GitHub/Release path). Compatible with DSH >=0.2.1-rc.2 <0.3.0; optional Python enables RapidOCR and Whisper. MIT licensed.
Read MoreTrack Diagnostic Analysis System: The Learner Navigation Skill for Standardizing Entry Points
The "Track Diagnosis Analysis System" (赛道诊盘仪分析系统) is a student assistant Skill published on SkillHub by community developer user_201a602c. Version 1.0.3, categorized under data-analysis, it does not require an API key configuration (`requires_api_key=false`). Targeted at student communities of the Nine-Dimensional Technical Analysis System (Yanming System, Lock Position Quantitative System, Nine Major Indicators), it addresses repetitive questions in bots regarding "where is the tool entrance," "how to use it," and "how to interpret indicators." When students input the trigger word "1", mention "Lock Position Quantitative System," "Nine-Dimensional Technical Analysis System," "Nine Major Indicators," or "Yanming," or when they inquire about the diagnosis tool's entrance, link, or instructions, the Skill directly outputs the tool link (https://saidao.yanming.site) and the instruction link (/guide.html) along with a brief one-sentence explanation. It can also explain indicator meanings, three-dimensional voting logic, and how to read reports. Upon request to analyze a specific A-share stock or generate a technical judgment report, it provides an online analysis entrance and can generate self-contained HTML report files locally and offline. The companion webpage requires no registration or password and includes Real-time News (Today's Sentiment, Sector Analysis, Today's Flash News, updated daily at 6:00, 8:00, 11:00, 15:00, 20:00, and 23:00), Track Analysis (generates reports in 10–30 seconds after selecting an industry, including key barriers, leading stocks, and hidden beneficiary stocks), and a Report Library (search and delete functions). Market data is sourced from East Money (东方财富), and news interpretation is AI-assisted. It is for learning reference only and does not constitute investment advice; the Skill itself serves solely for navigation and does not recommend stocks. No official install command is provided; it is distributed according to general SKILL.md specifications with the official address serving as the entrance.
Read Moredsh-memory-tree: Hierarchical SQLite Long-Term Memory with AST Indexing
DSH-Memory-Tree is a DSH memory plugin maintained by qinghuaju that layers AST node paths, embeddings, and hot/warm/cold SQLite storage. It captures conversations in the background and injects bounded memory-recall before each turn. Based on the GitHub README and package.json, this article describes its positioning, mem_* tools, the install.mjs installation steps, credential and Cordis configuration essentials, cross-session validation approaches, permissions, compatibility, and fallback considerations.
Read More