College Application Consultant – Industry-Major Matching: Reverse Engineering Major Selection Based on Industries
Introduces the SkillHub community skill “Gaokao College Application Advisor – Industry/Major Match” (@user_a76262b1/career-major-match, v1.0.0, category: education). The skill is triggered when users ask career-planning questions such as “what major should I study if I want to work in the XX industry” or “which majors are recommended for the XX field.” By analyzing the educational backgrounds of top professionals in an industry, it outputs a data-driven reference report on suitable majors. requires_api_key=false. The article explains its positioning, verifiable core functions, installation and activation methods, typical example questions, and applicable limitations. It also points out risks including the discrepancy between a rating of 100000 and 0 installations, as well as the undisclosed data source and industry coverage.
Read MoreDSH Task Timer: Bottom-of-Conversation Work Duration Pill and Sidebar Total
Introducing dsh-task-timer (1.6.0), a DSH client plugin maintained by ShuaihuiMou. It displays the work duration in the composer dock after deducting waiting time for tools such as ask_user_question, and shows a popup with the elapsed time from conversation creation to the present. The sidebar summarizes the total duration and tokens of all conversations, and fills in workspace session times as "N ago" relative timestamps. It explains the llmMs + toolMs − awaitMs measurement, the installation command, and usage notes.
Read MoreDeep Learning for Long Videos (Watching Long Videos like Reading Books): Turning Long Videos into Searchable Notes
Long-Video Deep Study is a community-based Skill following the general SKILL.md format, located at @user_d0c0b7f4/long-video-deep-study, version 0.2.0, category knowledge-management, publisher user_d0c0b7f4, source community, and tag requires_api_key=false. It targets high-density long-form videos such as interviews, courses, and video podcasts, and operates through six stages: downloading the video, transcribing it into timestamped SRT files, using AI to explain difficult sections, locating the original footage via keywords, generating bilingual (Chinese-English) translations, and exporting notes to a memory bank. Trigger scenarios include in-depth research on long-form videos, dissecting podcasts, and converting speeches into searchable notes. The official page does not provide installation commands, directory structure, or configuration fields, differences among the three types of clients also cannot be verified, and it is necessary to refer to the official address for accurate information. When using it, note that the process is relatively long, transcription quality depends on the audio, notes need to be stored in a memory bank, and the functional boundaries are only based on the official introduction.
Read MoreAgent Self-Evolution System: Enabling Reviewable Task Execution and Iterable Prompts
Most Agent Skills lack the ability to accumulate experience once their prompts are hardcoded: running the same task twice does not guarantee a better result the second time, nor are failure causes and effective prompt fragments recorded. Addressing this pain point, the Agent Self-Evolution System (@user_bae0772d/prompt-optimizer-bot, v1.0.0, categorized under SkillHub's ai-agent) offers three capabilities: intelligent task execution review, automatic prompt optimization, and capability iteration tracking. Together, these form a pipeline where the review generates the basis for optimization, the optimization produces modifications, and the records ensure traceability. This Skill is tagged with requires_api_key=false, meaning no additional API Key configuration is needed. This article faithfully documents the verification status: at the time of writing, two attempts to scrape the official page failed, so the installation commands, configuration fields, and official examples could not be verified; therefore, no speculative commands are provided. The "highlight" data provided in the task description (Rating: 99381, Downloads: 803) was inconsistent with the data scraped from the page (Rating: 100000, Downloads: 860, Installs: 0), and the article adopts the page data. Furthermore, it outlines applicable scenarios (e.g., having iterated prompts multiple times but lacking a clear record of each change, or needing to archive capability evolution) and points of caution (it is a review and optimization tool, not an execution engine; v1.0.0 has zero installs, resulting in few reference cases; review quality depends on the completeness of input records; and iteration records require manual planning for cleanup).
Read Moredsh-beme: Enabling Long-Term Memory for DSH Agents with Block Stores
**dsh-beme** is a memory-based plugin for **DeepSeek Harness** developed by **shua-781**. Upon installation, it provides `memory_remember` and `memory_recall` capabilities. It stores "events" (rather than entire chat logs) as a block-based repository in the local `~/.beme` directory, using streaming injection of relevant blocks to optimize context usage. This article, based on the official README, introduces the block repository concept, the installation command (`dsh plugin --profile web add github:shua-781/dsh-beme`), acceptance testing procedures, data directories, and uninstallation notes. It also notes that the plugin runs with DSH process privileges and that the data is stored in an unencrypted format.
Read MoreAI Large Model Expert | Model Service SLA Monitoring | AI-HIVE: Unifying Model Service Metrics into SLA
This Skill is published on SkillHub by 极睿科技 (Jirui Technology) AI experts at @infimind/ai-hive-gateway-model-sla-monitoring, version 1.0.0, source: enterprise, category: it-ops-security, tag: requires_api_key=true. It unifies availability, latency, quality, task completion, and support response into a single SLA, and provides routing policies, SLO metrics, fault-injection cases, observability fields, alert thresholds, and recovery runbooks. It also queries AI-HIVE at runtime for a current model, parameter, and pricing snapshot. Before execution, you must sign in and bind AI-HIVE remote MCP; paid, batch, external publishing, and production traffic switchover require separate confirmation; do not mask failures with unlimited retries; after an async task times out, resume by querying the original taskId to avoid duplicate submissions and duplicate billing. The article describes the scope of capabilities, prerequisites for enabling it, the official entry points, and typical usage, and notes limitations that the official page cannot verify installation commands or client differences.
Read Moredsh-mp4: Preview MP4 / M4V / MOV in DSH Sidebar
dsh-mp4 is a DSH client plugin maintained by shshjjj456. It previews MP4, M4V, and MOV files in the right sidebar using the browser's native player, replacing the "cannot preview" placeholder. The same package registers both dsh-better-sidebar and native document preview, supporting Desktop and dsh web. To install, search for dsh-mp4 in the plugin marketplace or run `dsh plugin --profile web add dsh-mp4`. Requires Node≥20, dsh≥0.2.1-rc.5, MIT license.
Read MoreOne-Person Company (OPC) Policy Expert: Consolidating Policy, Financial & Tax, and Legal Issues into a Single Skill
"One Person Company Policy Expert" (OPC) is released by user_54323d85 (Version 2.0.0, Source: community, Category: professional). It integrates nine categories of capabilities into a single Skill: tax law consultation, legal regulation inquiry, financial and tax case retrieval, China market policy inquiry, in-depth web research, multi-engine search, social security and housing fund inquiry, financial stewardship (bookkeeping/invoices/taxation/reports/cash flow/Golden Tax Phase IV self-inspection), and legal risk control (contract review/templates/compliance checkups). It targets policy, financial, tax, legal, and compliance issues relevant to one-person companies and individual industrial and commercial households. The official designation indicates `requires_api_key=false`. This article explains the capability groups, official address, and SkillHub page, providing reproducible question examples. Since the retrieved materials lacked CLI installation commands and details on client differences, the installation section only lists verifiable entry points, advises referring to official pages, and suggests verifying policy conclusions against the latest documents. Furthermore, major matters should be guided by opinions issued by licensed professionals.
Read MoreDSH Prototype Annotator: Click elements in the Desktop sidebar to hand modification requests to Agent
dsh-prototype-annotator is a DSH Desktop client plugin maintained by shiyoungwoo that provides Codex-style click-and-select page annotation. It opens the local frontend in the sidebar Browser, where users can select elements and write modification requests. Multiple requests are merged into an "N annotations" tag, sending context such as DOM selectors, styles, and viewport along with the message. The plugin does not directly modify source files and requires no additional services. Installation can be performed by instructing the Agent to execute `install_bundle` via conversation, or by running `dsh plugin --profile desktop add github:shiyoungwoo/dsh-prototype-annotator`. It is compatible with macOS Desktop starting from 0.2.1-rc.3 (package declaration: dsh >=0.2.1-rc.3 <0.3.0), while Windows and Linux support remain unverified. Licensed under MIT 0.0.3.
Read MoreHolistic Business Operations Expert: Translating Full-Scale Consulting into Executable Templates
The Business Full-Solution Operations Expert is a general-purpose Skill.md-format skill on SkillHub, released by Chanmeng Technology. It is version 1.0.0, categorized under business-ops, sourced from enterprise, and tagged requires_api_key=false. The page displays a score of 100,000, 916 downloads, and 0 installations. It provides comprehensive business consulting and execution templates covering business model diagnostics, sales presentation-driven closing, investment attraction and financing, equity design, organizational systems, partnership mechanisms, and traffic growth. It solidifies common business full-solution modules into reusable structures, making it suitable for developers who need standardized business full-solution responses. Installation is done by obtaining it through SkillHub and adding it to the Skill directory. Official documentation does not provide differentiated installation commands for Cursor, Codex CLI, or Claude Code, nor does it offer directly reproducible invocation examples. When using it, note that it provides consulting frameworks and templates, not legal, financial, or investment advice. Before implementation, it should be adjusted based on your industry and stage, and reviewed by professional consultants.
Read Moredsh-openbot-artist-bridge: Connecting OpenBot Conversations to the Existing DS Bot in DSH
The DSH client plugin `dsh-openbot-artist-bridge`, maintained by shiyoungwoo, connects OpenBot's ACP frontend `acp_bridge.py` with `runtime.mjs` in DSH via an AF_UNIX socket, forwarding conversations to existing DS Bot sessions while sharing the same identity, memory, and workspace. Compatible with DSH 0.2.1-rc.3 to pre-0.3.0, Node 22.19+, Python 3.9+, and macOS/Linux. Installation is achieved via `pnpm add` with a link or GitHub reference; after configuring `botId` and `socketPath`, register the Python bridge command in OpenBot. Released under the MIT License, version 0.1.0.
Read MoreData Parsing, Structured Extraction, Information Validation: Transform Arbitrary Text into Structured JSON with Confidence Scores
This skill addresses the pain point developers face when dealing with unstructured text: models can output text, but field formats are inconsistent, missing values are undetected, and downstream systems cannot process them. It consolidates text parsing, field extraction, and validation into a single step: inputting any text generates structured JSON output, with confidence scores for each field and a list of missing fields, enabling downstream systems to decide whether to use the results directly or route them for manual review. This skill is based on the universal SKILL.md format, located at @user_2fd890c9/ape, maintained by community publisher user_2fd890c9. It is in version 1.0.3, categorized under "data-analysis", and its tags indicate it does not rely on external API keys. The provided material lacks specific installation commands, configuration fields, and reproducible examples. This article only elaborates on verified information and highlights the need for self-verification regarding its community origin, limited downloads, and installations.
Read MoreAike SCRM - WeCom Assistant: Entrusting WeCom Customer Data Query and Management to the Agent
This article introduces the "Aike SCRM - Enterprise WeChat Manager" Skill on SkillHub (@user_82d36ed4/2026aike-scrm-skills, Version 1.0.10, Source: Community, Category: Business Ops, Published and maintained by user_82d36ed4). Designed for Enterprise WeChat customer operations, it enables Agents to query or manage five categories of business data within Aike SCRM (www.ikscrm.com) during conversations: customer information, customer tags, enterprise tags, customer groups, and group members. This eliminates the overhead of writing scripts for individual data fetches and repeatedly consulting documentation. The trigger conditions are broad: the Skill activates whenever a user is engaged in Enterprise WeChat customer operations and management scenarios, even if Aike, SCRM, Open Interfaces, or APIs are not explicitly mentioned. Therefore, it is necessary to clearly demarcate its boundaries in the Skill description compared to other customer-related Skills in the workspace. The provided material does not include official installation commands, directory conventions, or configuration file formats; only the tag indicates requires_api_key=true. Consequently, this article provides no commands but instead explains the integration approach and points out that attention should be paid to official pages. The article concludes by providing the official website and contact pages, reminding users that this Skill is sourced from the community, and emphasizing the need to verify capability scope and data definitions before deploying it in production.
Read Moredsh-classic-mode: One-click Open DS Bot Web Configuration in Classic Sidebar
dsh-classic-mode is a DSH client plugin that adds a DS Bot entry to the bottom of the classic Agent sidebar. Upon clicking, the host starts or reuses web profiles (such as 'classic'), parses token-bearing URLs, and opens them in Safari, eliminating the need to manually start `dsh web` or paste addresses. It requires configuration of `cliPath`, `profile`, `port`, and `browser`; it also includes cleanup for lingering web processes and security notes. Installation is performed via `pnpm add github:shiyoungwoo/dsh-classic-mode` followed by applying the `cordis.patch`.
Read MoreMarket Watch Bot: A-share Market Monitoring & Multi-indicator Alerting Skill
The "Market Watch Robot" is a community-oriented Skill for A-shares, maintained by user_1037892e (version 8.0.2) and based on the universal SKILL.md format. It retrieves real-time quotes via the Sina online API, computes five core indicators—MACD/KDJ/RSI/BOLL/MA—and delivers v6.0 knowledge enhancements (ATR/ADX/volume ratio/gap classification/20-day valid breakout/baseline-point dynamic stop-loss), v7.0 Wyckoff Volume Price Analysis (VPA) system (volume-price divergence/key candle volume verification/false breakout filtering/VAP support-resistance/multi-timeframe linkage), and v8.0 network resilience (exponential backoff retries, friendly diagnostics, one-click self-check). It integrates a multi-indicator rule engine and bullish/bearish signal statistics, and supports automatic alert push via DingTalk Webhook. The article details features, installation/configuration, and typical usage, and notes caveats such as data-source stability and that signals are not investment advice.
Read MoreCase Analysis and Group Interview Simulator: Bringing Case Interviews and Group Interview Practices into a Repeatable Training Environment
A general SKILL.md-format Skill for interview preparation targeting consulting, product, and strategy roles, located at @user_fa14b5bc/case-interview-simulator, version 1.0.0, from a community source, categorized under education, with requires_api_key=false. It features built-in single-person case breakdown and no-leader group mock interview modes, a library of 7 types of business cases, analysis framework guidance, a follow-up question simulation engine, and a five-dimensional scoring system. The provided materials do not include installation commands or reproducible examples, and the official page could not be retrieved. The body text only includes verifiable parts. The page shows a rating of 100000, 863 downloads, and 0 installs.
Read Moredsh-model-avail: Probe model availability with real requests and reorder the selector
dsh-model-avail is a DSH plugin maintained by Shino113-1 (v0.3.1, MIT). It sends a minimal request containing a ping and a single output token to each model, categorizing the results as available, unavailable, or unknown. It caches this status for 6 hours and annotates providers with "available/total" counts in the Web UI model selector, sorting them in descending order of availability. Model names are color-coded with green, red, or gray to indicate their status. Installation can be performed by entering `github:Shino113-1/dsh-model-avail` into the plugin panel or by running `scripts/install.mjs`. This requires Node.js ≥ 20 and a subsequent restart of DSH. The plugin provides HTTP interfaces such as `GET /dsh-model-avail/state`; the detection process has limits on concurrency and requests per round, and billing-sensitive users can adjust the relevant constants.
Read MorePonytail Coding Standards Skill Set: Enforcing Code Standards via Agents
The "Ponytail Code Standards Skill Set" is an AI Agent skill package published by the SkillHub community (ID: @user_3c6cb52e/ponytail-png, v1.0.0, Category: dev-programming). It contains six internal sub-skills designed to enable Agents to execute coding tasks according to established standards. It is compatible with Claude Code, Codex, and WorkBuddy, and requires no API key (requires_api_key=false). This article first introduces the skill set and its maintainers, followed by an explanation of its verified core capabilities. Since the public page lacks detailed names of the sub-skills and copyable installation commands, the installation section provides only general, coordinate-based activation steps, avoiding any fabricated commands or configuration fields. Similarly, the "typical usage" section offers only suggestions for invocation methods, with clear notations where official examples are missing. Finally, the article outlines the target audience and key considerations: as this is a community-published personal project with 0 installations, 907 downloads, and a rating of 100000, it is recommended to test it on non-critical projects before adoption and to verify whether its standard coverage aligns with your team's requirements.
Read Moredsh-optimization-2027: Single-Session Inference Optimization Bundle
dsh-optimization-2027 is a DeepSeek Harness bundle maintained by shiliuyu316. Within a single session, it provides five switchable, tier-based reasoning protocols, a conclusion ledger, declarative PowerShell tools, a persistent shell, runtime context re-injection, a read-only analyst, and an input box side panel. The tiers inject only English-language rules and do not modify model parameters such as reasoningEffort; in the off tier, protocol overhead is zero characters. Based on the official README, the article covers installation via plugin_manager install_bundle, the /dopt2027 user command, Cordis configuration options, and data directories, and describes token cost accounting conventions. It also cautions that the plugin runs with the DSH process’s privileges and notes Windows-first validation as a compatibility limitation.
Read MoreWeChat Official Account Layout Beautification: 7 Industry Templates with Ready-to-Paste HTML
A typesetting skill designed for WeChat Official Account (OA) authors, located at @user_f3268e7b/wechat-article-beautify, published and maintained by community developer user_f3268e7b. It is categorized under content-creation with a community source, currently at version 1.0.1, and requires no API key. By inputting completed Markdown, plain text, or HTML articles, users can apply 7 built-in industry templates (Business Consulting, Education & Training, Finance, Gourmet Lifestyle, Healthcare, Emotional Literature, and Technology & Internet) to generate finished HTML with fully inlined styles, ready to be pasted directly into the WeChat editor. Supported features include applying a single template, batch applying all templates, and side-by-side comparison of multiple versions. Typical triggers include phrases such as "beautify article," "article layout," "apply template," "reformat," or "WeChat OA typesetting," or when users request reformatting by templates, generating multi-version comparisons, or producing pages compatible with the WeChat editor. Note: Installation commands and configuration fields could not be verified from the official page and are disclosed as such in the text.
Read MoreDamodaran's Valuation Philosophy: Breaking Down Intrinsic Value Assessment into Ten Steps
The Damodaran Valuation Philosophy is a professional-classification Agent Skill (version 1.0.0) published by community developer @user_231765ee on SkillHub, tagged with requires_api_key=false. It encapsulates Aswath Damodaran's valuation methodology into a ten-step workflow covering story construction, intrinsic value assessment, discounted cash flow analysis, relative valuation, scenario analysis, and decision matrices, enabling Agents to progress from qualitative judgment to final investment decisions rather than directly providing target prices. This article introduces the framework's capabilities and limitations, installation and activation procedures, typical invocation examples, and usage considerations. During retrieval, the official page was unreachable; installation commands should be verified in practice, and the framework itself does not provide real-time financial data.
Read Moredsh-mcp-manager: Managing MCP Servers in DSH
A community plugin, dsh-mcp-manager (shiliu153), for DeepSeek Harness users centrally manages MCP entries in the ~/.dsh/cordis.patch.yml configuration file. Prior to writing, it validates entries according to rules from the DSH loader and @deepseek-ai/dsh-mcp-client, ensuring atomic writes with backups. It provides nine mcp_manager_* agent tools, a zero-dependency CLI, and a web dashboard, with marketplace integration with the official MCP Registry. Features include copy-and-paste installation across four formats (npm, GitHub, URL, JSON), real handshake probes, start, stop, and delete operations, and credential rotation (the dashboard does not write !!js). The installation command is plugin_manager install_bundle accompanied by the GitHub URL.
Read Morebetter-plugin-combine-dsh-society: DSH Plugin "General Office" Read-Only Directory
DSH plugins are independent entities, making it difficult to quickly determine their ownership, configuration locations, and historical records. The `better-plugin-combine-dsh-society` (Office) plugin provides a read-only local directory that aggregates plugin functions, tools, configuration entries, data directories, and memory entry counts. It utilizes `office_catalog` and `office_ask` for on-demand indexing and cross-`meow-memory` retrieval, ensuring it does not write to the database, modify configurations, or execute tasks on behalf of the user. Based on the official README, this article covers installation (`plugin_manager install_bundle`), configuration options, and known limitations, making it suitable for DSH users managing numerous plugins.
Read MorePharmaceutical Financial Digitalization Product Manager: Translating Opportunities into Deliverable BRDs/PRDs
A Skill targeting financial digitalization scenarios for pharmaceutical enterprises, located at @baitongai/pharmaceutical-financial-digital-product-manager, published by the Baitong AI service platform (version 1.0.0), categorized as business-ops, sourced from enterprise. Its capabilities include opportunity identification, solution design, BRD/PRD authoring and Word-based deliverables, with 8 pharmaceutical industry-specific scenarios and 8 general financial scenarios built in. It supports dynamic .docx generation, project input-output estimation, ROI quantification calculations, and intelligent scenario recommendation. The tag requires_api_key=false indicates that invocation does not require an additional API Key. Official materials do not provide command-line installation instructions or configuration fields; the main content only supplies the coordinate and the official URL, and notes that there are no publicly differentiated steps for Cursor / Codex CLI / Claude Code. It is also flagged that: industry scenarios are bound to the pharmaceutical sector, installations remain at 0 with low usage, the scenario list and template details have not been disclosed, and generated outputs require manual review.
Read MorePorter's Five Forces Analysis: Running Industry Structure Analysis with an Agent
Manual assessment via the Porter's Five Forces model is time-consuming and prone to oversights. Released by Engicool (@engicool/porters-five-forces-skill, version 1.0.0, category: professional, tag: requires_api_key=false), this Skill guides users through a step-by-step evaluation of the five forces: supplier bargaining power, buyer bargaining power, threat of new entrants, threat of substitutes, and competitive rivalry. It generates assessments of industry attractiveness and the competitive landscape, making it suitable for market entry decisions, competitive strategy formulation, and investment portfolio evaluations. The manufacturing, supply chain, and quality-related industries are typical application domains. This article describes the Skill's positioning, core features, activation information, typical usage, and applicability limitations. Since official materials do not provide installation commands or platform-specific activation methods, the article states this explicitly and recommends referring to the SkillHub page for accurate instructions.
Read More