dsh-usage-replay: Cross-Session per-Model Replay of Token Usage
dsh-usage-replay is a model inference plugin developed by Eclipse-NotFound for DeepSeek Harness. It replays session logs in ~/.dsh/sessions, aggregates provider-reported token usage by provider/model, by day and workspace, and provides a sidebar usage Tab, per-model chips in Composer, and two local read-only HTTP APIs. It only counts tokens, with no cost estimation or network requests. Requires DSH 0.2.1+; the usage Tab optionally depends on dsh-better-sidebar ≥0.28. Install: dsh plugin --profile <profile> add dsh-usage-replay.
Read Moreverified-search-pro: Adding an Information Quality Verification Layer to AI Agents
verified-search-pro (VSP) is a community-published quality assurance layer Skill, version 2.1.0, categorized under data-analysis, which requires API Key configuration. It integrates intelligence methodologies including OSINT, ACH, SIFT, and the Admiralty dual-axis rating, and aggregates multiple search engines such as Tencent Cloud WSA, Baidu Qianfan, Tavily, Bing, Sogou, and DuckDuckGo. It evaluates search results and assigns A–E confidence levels based on source reliability, content credibility, the number of independent confirmations, and disconfirmation status. It outputs both human-readable Markdown reports and machine-readable claims-json/evidence-pack files. VSP supports two modes: configuring full APIs as the search foundation, or using the --input-results parameter in quality inspector mode to verify results obtained by the host. It is well-suited for in-depth research, fact-checking, policy tracking, and competitor background checks.
Read MoreMain Image and Banner Expert: Full Process of E-commerce Main Image and Banner Design Skill
The "Key Image Banner Expert" is a SKILL.md format Skill designed for e-commerce key images and banner design. Released by community publisher user_442551c3 (Version 1.0.0), it is categorized under design-media and requires no API Key. It covers platform specifications for Taobao 800x800, Tmall, JD.com, Pinduoduo, and Douyin E-commerce, along with the 3-second "Visual Hammer" rule, the "Three-Eye" information hierarchy, color psychology, text ratio control (under 30% to prevent violations), strategies for scenario-based images and video key images, A/B testing methodologies, differentiated design, and multi-platform banner size adaptation. Triggered by keywords such as main image, banner, and CTR, the article introduces core functions, installation and activation methods, six typical usage scenarios, and important precautions.
Read MoreLitigation Visualization Skill: Convert Case Work Papers to PPTX or Offline HTML
The Litigation Visualization Skill, published by user_7d77e6da on SkillHub (Location: @user_7d77e6da/litigation-visualization-zh), is Version 1.0.3, categorized under professional, and sourced from the community. It processes frozen 07 case working papers to generate litigation visualization PPTX or offline HTML review drafts. It can directly handle real-world cases, accepting both real-name and anonymized materials without prior authorization or anonymization. It only halts when the 07, TEXT PASS, A01-A09 anchor set, or hash does not close. Additionally, it can replay a fixed three-event Control/A/B and a three-page draw.io composite demonstration, using only built-in samples. The tag `requires_api_key=false` indicates that no API key configuration is required for invocation. The official URL is https://api.skillhub.cn/user_7d77e6da/litigation-visualization-zh. The retrieved documentation does not provide installation commands, command-line entry points, or parameters; this article will not supplement them, and they must be confirmed on the official page before integration.
Read Moredsh-notes-wall: DSH Sidebar Sticky Notes Wall, Local Quick Notes
EchoTint's client plugin dsh-notes-wall provides a notes wall in the DSH sidebar: 240 ms automatic save, six paper colors with pinning, filtering via in-body #tags, search, and JSON import/export. Data is stored only in localStorage (dsh.notes-wall.notes.v1), and Host apply is left empty. Install via `dsh plugin add https://github.com/EchoTint/dsh-notes-wall`; a DSH restart is required after installation or after modifying client.js. MIT 1.1.2.
Read More3-Point Technical Indicator: Converting Oversold Golden Crosses and Overbought Death Crosses into Reusable Technical Signals
“3-point technical indicator” is a community-sourced technical analysis Skill, located at @user_63c06e28/3-percent, published and maintained by user_63c06e28, version 1.0.1, classified as professional, and does not require an API Key. It is implemented based on TongDaXin's SKDJ variant and is used to identify golden cross buy signals in the oversold zone and dead cross sell signals in the overbought zone. It is designed for users who need to incorporate technical buy/sell rules into their development workflow, and can reduce the repetitive work of re-implementing indicators from market data and threshold judgments each time. The call approach is to input A-share or ETF market data, obtain indicator values, compare them with the oversold and overbought zones to determine golden cross or dead cross, and use the golden cross as a buy signal and the dead cross as a sell signal. Since the official page was temporarily unavailable at the time of writing, the installation command, target directory, and configuration fields were not verified, and the main text did not provide unverified commands; also, this output is only a technical signal, threshold differences will affect the conclusion, and it does not constitute a guarantee of price rise or return.
Read Moredsh-mijia-sidebar: Mijia Control Panel in DSH Sidebar
dsh-mijia-sidebar is a DSH client plugin maintained by EchoChan314 that provides a Mi Home (Mijia) control panel in the sidebar. Device, property read/write, action, and manual scene operations are invoked from the Host half via the /mijia RPC, which calls the local mijiaAPI CLI without consuming model context. The article covers prerequisites (Python ≥ 3.10 and mijiaAPI login), official installation methods such as `plugin_manager install_bundle`, the configuration file `mijia-sidebar.config.json`, typical panel usage, and troubleshooting common error codes. It also notes that the plugin runs with DSH process privileges and advises reviewing the source code and MIT license before installation.
Read MoreCybersecurity Self-Study Coach: An Organized Roadmap and Self-Rescue Strategies for Learners
Common challenges in self-directed cybersecurity learning include scattered learning paths, disordered hands-on practice, and encountering roadblocks with no one to ask. Published by Kunlun Growth (located at @kunlungrowth/awesome-cyber-security-university, version 1.0.0, categorized as "education" on SkillHub, requires_api_key=false), this Skill organizes learning along a structured roadmap progressing from foundational networking and operating systems to Web security, reverse engineering, and cryptography. It provides signals for practical exercises and self-rescue strategies for learning bottlenecks, emphasizing legal and compliant cyber range training. The ultimate goal is to cultivate a security mindset rather than merely accumulating tools. Since the documentation lacks specific installation commands and reproducible examples, users should verify the details on the SkillHub page prior to adoption.
Read MoreBidding: Breaking Down Tendering and Government Procurement Processes into Traceable Auxiliary Workflows
"bidding" is a Skill designed for the Chinese context, covering the full workflow of tendering, bidding, and government procurement. Published by the community publisher **user_7c6357d4** (version 1.0.0), it is categorized under **business-ops** with the tag **requires_api_key=false**. It breaks down the entire process into sequential stages with defined deliverables, ranging from tender notification discovery, procurement needs investigation, document drafting, and fair competition review, to PDF/Word/scan parsing, extraction of qualification and disqualification criteria, scoring methodology modeling, bid decision-making, chapter-based generation of hundreds-of-page bid documents, requirement-evidence-response matrices, objective score verification, subjective score assessment assistance, price score simulation, compliance review, Q&A clarification, pre-submission checks, bid opening/evaluation assistance, contract alignment, performance acceptance, and post-project review. Unlike approaches that feed entire documents to a model for full-text summarization, "bidding" ensures that the output of each step becomes the input for the next, creating a traceable chain. The official introduction explicitly states that the AI provides only auxiliary analysis and does not replace the primary responsibility of the procurer, the independent review by the bid evaluation committee, legal opinions, or manual signing and submission; therefore, content involving bid rejection, clarifications, and contract terms still requires manual confirmation. As the official and SkillHub pages were inaccessible during the writing of this article, details regarding installation commands, directory locations, configuration fields, and editor-specific activation differences were not included. Only version, source, category, and tag information verifiable from retrieved data were retained. Readers are advised to independently visit the official address to verify the publisher, version, and download count before deciding whether to adopt the tool.
Read MoreTravel Culinary Assistant: Cuisine Retrieval and Dining Planning in Travel Scenarios
# Introduction When adding travel capabilities to an Agent, a common practice is to let the model directly answer "what is good to eat in [location]". However, in actual execution, numerous issues arise: answers fluctuate with the model, sources are lacking, restaurants near tourist attractions are mixed with local specialties, and catering arrangements for multi-day itineraries are impossible to handle. The Skill introduced below converges such requirements into a defined capability—destination food retrieval, nearby restaurant recommendations, local specialty identification, and multi-day dining planning. # What is this The Travel Food Assistant is a Skill for food recommendation and dining planning designed for travel scenarios. It is published by the community account user_b95ee7e5, located at @user_b95ee7e5/travel-food-guide-new. The current version is 1.0.1, and the source is marked as
Read Morehy-memory4dsh: Integrating Session, Project, and Personal Three-Layer Long-term Memory for DSH
hy-memory4dsh (dsh-hy-memory), maintained by Echo-Smith, implements three-tier memory—personal, project, and conversation—within DSH 0.2 preview through the Hy-Memory 2.0 SDK sidecar. It supports chat/working/coding scenarios, automatic recall and capture, a Settings memory panel, and MemHub interoperability. This article introduces the mechanisms, installation commands, configuration options, and usage considerations based on the official README, with storage located at $DSH_HOME/integrations/hy-memory/.
Read MoreFinXData: A Skill for Querying A-Share, Hong Kong Stock, and Commodity Data
# Preface When building financial Agents, acquiring market data is often the first hurdle: A-shares, Hong Kong stocks, commodity prices, financial reports, and macro indicators each correspond to a separate set of APIs, most of which require individual key registration and quota management. FinXData is a community Skill that consolidates these queries into a single entry point, with the official coordinate @user_8ef78c30/finxdata. # What is this FinXData is an Agent Skill based on the universal SKILL.md format, published and maintained by the community publisher user_8ef78c30. It is tagged with the source "community", categorized as "professional", and is currently at version 1.0.17. The problem it solves is: when a user wants to query stock prices...
Read MoreInsight into Individual Stock Major Players: Diagnosis of Smart Money Status
"Individual Stock Main Force Perspective" is a deep diagnostic Skill for individual stocks, located at `@user_6c1a7171/geguzhulitoushi`, version 1.0.0, released by community developer `user_6c1a7171`. Classified as "professional" on SkillHub, it requires no API Key configuration. It is used when users provide a 6-digit stock code and inquire about the "main force" (market maker/smart money) situation. It identifies who is controlling the market (hot money vs. institutions), determines the smart money's phase (accumulation, markup, shakeout, or distribution), assesses current trading opportunities, and calculates entry, stop-loss, and take-profit points. It is activated by trigger words such as "个股主力透视", "透视一下 600869", "谁在控盘", and "吸筹还是派发". The retrieved data does not include installation commands; it can be enabled via the official address or the SkillHub page using the coordinates. Note: This only handles individual stocks; for batch requests like "today's main force list," use `ashare-mainforce-identifier` or `ashare-mainforce-scan`. Diagnostic conclusions do not constitute investment advice.
Read MoreCustomer Acquisition Assistant: Skill for Filtering Leads from Douyin Comments by Keywords
The Customer Acquisition Assistant (@user_fb71c45d/huokezhusho, v1.0.0, community source, business-ops category, SkillHub rating 100000, 1187 downloads) is a community-published Agent Skill designed for customer acquisition in Douyin's public domain comment sections. The fixed workflow consists of two steps: first, searching for videos by keywords to obtain the aweme_id, and then using the aweme_id to retrieve comments and extract the nickname, uid, sec_uid, region, and comment content. It transforms the manual process of scanning comments to identify high-intent users into a reproducible pipeline, making it suitable for keyword-driven lead mining, intercepting traffic in competitors' comment sections, and preliminary user segmentation before private domain traffic diversion. An API Key is required (requires_api_key=true). As the provided documentation lacks installation commands, configuration fields, and reproducible examples, this article does not elaborate on them. It is recommended to conduct small-scale validation in a test environment to verify the completeness of API responses and field consistency, while paying close attention to the compliance requirements regarding the processing of personal information in comment sections.
Read Moredsh augment: Token Compression and Evidence-Driven Self-Evolution for DeepSeek Harness
dsh augment is a DeepSeek Harness plugin (v0.0.1 preview) maintained by Ebony-Vinyl that compresses tool outputs and collapses duplicate reads without modifying model weights, while making per-request costs transparent via a token ledger and a tool-schema overhead panel. It supports CCR externalization with augment_retrieve retrieval, project maps and environment experience, and loop protection. Optional features include a self-evolution workspace and a Capability Gym, with adoption gated by statistical evidence. Installation requires running `node scripts/install.mjs` in the plugin directory after shutting down DSH; data is stored in `~/.dsh/dsh-augment`. MIT licensed.
Read MoreGraham Securities Analysis: Turning the valuation methods from the 6th Edition of *Security Analysis* into a Callable Skill
This article introduces the "Graham Security Analysis" Skill on SkillHub (@user_5d9e7cf1/graham-security-analysis, version 1.0.2, categorized as professional, sourced from the community, requires_api_key=false). It breaks down the analytical framework of Graham's *Security Analysis*, 6th Edition, into executable steps, covering intrinsic value, margin of safety, the distinction between investment and speculation, four principles of fixed-value investment, income statement red flags, the five-step balance sheet method, net working capital and liquidation value, along with three supplementary materials: a quick reference sheet, conversation examples, and FAQ. The aim is not to memorize famous quotes, but to learn how to weigh securities using financial statements. The article outlines the Skill's positioning, five core functional groups, typical usage at the intent level, applicable scenarios, and four limitations, and candidly notes that the official installation command and directory could not be verified in the retrieved materials and should be subject to the SkillHub page.
Read MoreData Visualization MCP Server: A Toolbox for Delegating Chart Generation to Models
The Data Visualization MCP Server (@user_053d27d0/mcp-data-viz-zx, version 1.0.1, community release) provides 32 MCP tools covering bar charts, line charts, pie charts, scatter plots, heatmaps, box plots, histograms, radar charts, HTML reports, KPI cards, data tables, and dashboards. It transforms plotting from script-writing to direct tool invocation by the model, reducing back-and-forth costs in the data analysis workflow. It is categorized under data-analysis and tagged with requires_api_key=false. No official installation commands, configuration fields, or reproducible examples were found in the documentation; this article reports this as-is and refers to the SkillHub page for details.
Read Moredsh-programmer-chatroom: Decentralized P2P Group Chat Client on DSH
dsh-programmer-chatroom is a P2P group chat client provided by EaveBounty for the DeepSeek Harness web profile. It launches dshc-relay.js via a subprocess, turning each DSH machine into a mesh relay where messages are deduplicated by msgId and are not persisted. Nodes perform offline screening of sensitive content using locally embedded DFA and regular expression rules. It supports LAN UDP discovery and Internet rendezvous (Cloudflare Workers, cpolar). The recommended installation command is `dsh plugin --profile web add @eave_bounty/dsh-programmer-chatroom`, followed by a DSH restart. Since the plugin runs with the permissions of the dsh process, the source code and the MIT license should be audited before installation.
Read MoreAudit Flowchart Generator: Generating Internal Control Audit Flowcharts from Natural Language
This article introduces the SkillHub community Skill, the "Audit Flowchart Generator" (ID: @user_b53e245c/nigo-flowchart-generator, Version 1.0.0, Category: design-media, Source: community). Tailored for internal control and audit documentation scenarios, this Skill transforms natural language descriptions of business processes into flowcharts in four formats: .drawio, PNG, SVG, and VSDX. It supports various diagram types, including swimlane diagrams, approval workflows, expense reimbursement processes, internal control workflows, fund flows, and procurement processes. The Skill does not require users to explicitly state the word "flowchart"; it can be triggered by requests related to drawing, visualization, mapping business processes, or phrases such as "help me draw a [XX] process." It is marked with `requires_api_key=false`. According to the official page, its current rating is 100,000, with 1,064 downloads and 0 installations. The official documentation does not provide installation commands or configuration file fields, nor does it specify the enabling differences among Cursor, Codex CLI, and Claude Code. This article does not fabricate such details and recommends consulting the official page for the most accurate information.
Read MoreProject Knowledge Graph: Structuring the Relationships among Projects, Files, and Tasks
Project Knowledge Graph is a community Skill on SkillHub (@user_fa4109f3/summary-v2, v3.0.0). Belonging to the "knowledge-management" category, it was published by user_fa4109f3 and sourced from the community. It requires no API Key to use. This tool scans projects to structure the relationships between projects, files, and tasks into a knowledge graph. By integrating FTS5 full-text search, automated association analysis, and semantic search, it addresses issues such as opaque repository structures and the difficulty of tracing the correspondence between tasks and files. The article introduces its core features, installation entry points, typical usage, and use cases, while also highlighting potential concerns such as a failed official webpage crawl, unverified installation commands, and zero install volume.
Read MoreWorkBuddy Daily Point Auto-Claim: Setting up Daily Scheduled Clicks for the Client via pyautogui Template Matching
This Skill automates the process of claiming daily points in the WorkBuddy desktop client. It utilizes PyAutoGUI template matching instead of hard-coded coordinates to locate the entry point within the fixed UI and execute scheduled clicks. It includes a guided template calibration script, allowing users to verify the template area on their local machines to accommodate varying window positions, scaling settings, and light/dark modes. A troubleshooting checklist is also provided, covering common issues such as template regions, window z-order, matching thresholds, and multi-monitor offsets. Officially categorized under "office-efficiency" (Rating: 99066; Downloads: 955), it is triggered when automatically claiming WorkBuddy daily points or setting up scheduled click workflows within the client's fixed UI. As the official webpage scrape failed during this session, the installation commands, directory paths, and differentiated configurations for various AI tools could not be verified; thus, this description relies solely on the documented capabilities and usage concepts without fabricating official steps. This tool requires a relatively stable client UI and a graphical desktop environment; recalibration is necessary following any UI redesigns.
Read MoreYonyou U8+ Intelligent Assistant: Handling Document Operations for Six Major U8+ Business Scenarios via Agent
The Yonyou U8+ Intelligent Assistant (@user_d6971ddf/u8integratedskill, v1.0.3, community, business-ops) is a community-based Skill for U8+. It consolidates common operations from six major business domains—General Ledger, Accounts Receivable, Accounts Payable, Sales, Purchasing, and Inventory—into a single capability. It supports the creation, inquiry, and approval of documents such as sales orders, purchase orders, and inventory documents, while offering queries for pre-configured reports including Sales Statistics, Delivery Statistics, and Inventory Transaction Journals. Before integration, note the tag `requires_api_key=true`, indicating that an API Key must be configured. Furthermore, when querying pre-configured reports, the inquiry scheme must have the "[Support Mobile App]" option selected; otherwise, data retrieval may fail. As installation commands and configuration fields are not provided in the documentation and the official page retrieval failed, this article makes no speculative additions and only outlines the capability boundaries and usage prerequisites based on verified information.
Read MoreIntellectual Property Moat — One-Stop Management of Trademarks, Patents, and Copyrights: An Introduction to the Skill
## Introduction to the Intellectual Property Knowledge Base Management Skill for Legal Departments, Agencies, and Innovative Enterprises First, it addresses pain points: trademarks, patents, copyrights, and trade secrets are scattered across spreadsheets and documents, with searches relying on memory. Then, it provides positioning information: sourced from the community, located at @user_99b1d2cb/zhishichanquanguanli, published by user_99b1d2cb, version 1.0.1, categorized under knowledge-management, with a tag requires_api_key=false. Core functions cover five areas: trademark registration, patent application, copyright protection, trade secret compliance, and IP enforcement. It claims to be based on the TRIPS Agreement, PCT system, Paris Convention, WIPO Academy, and Espacenet. Installation and activation instructions only provide the official address and SkillHub page. Since materials do not include installation commands or configuration fields, no fabricated information is provided. It states that it follows the general SKILL.md format and can be loaded in hosts supporting this format. Typical usage includes: unified structure import, organizing requirements according to PCT or Paris Convention, verifying classification numbers through public entries, and separately grading trade secrets—emphasizing that it does not replace legal judgment. Applicable scenarios are provided for three types of teams. Notes indicate that ratings and download numbers are based on crawled materials (100000/1008), which differ from the clue's 99895/934. Manual review is required for community sources and changes in international application rules. The conclusion reviews its value and provides the official address.
Read MoreReal Estate Livestreaming Script: A Ready-to-Use Full-Process Broadcasting Skill
Property Live Streaming Scripts (@kunlungrowth/property-live-script, v1.0.0, Provider: Kunlun Growth, Category: content-creation, Source: enterprise, No API Key required) breaks down a live broadcast into four segments: opening warm-up, property presentation, interactive lead generation, and closing pitch. By utilizing trigger words (property live streaming, live streaming script, property presentation, lead generation script), it generates ready-to-read, step-by-step scripts, significantly reducing the effort required for hosts to improvise on the fly. The article covers its positioning, core features, installation and activation methods, typical use cases, and usage notes. It specifically notes that the tool only generates a process framework, meaning specific property details and promotional policies must be manually supplemented.
Read Moredsh-mimo-asr: Connecting DSH Go microphone to Xiaomi MiMo Cloud ASR
dsh-mimo-asr is maintained by Easyhoov and implements the speech/catalog and speech/transcribe Remote for DSH, connecting Go input box microphone transcription to the Xiaomi MiMo (mimo-v2.5-asr) cloud, eliminating the need for a local SenseVoice model. The installation command is `dsh plugin --profile <profile> add dsh-mimo-asr`, with API Key configuration supported via the App settings page, environment variables, and key files. Do not enable alongside the official voice input bundle. MIT licensed.
Read More