Computer Use: An AI Agent Skill for the General SKILL.md Specification
Computer Use is an Agent Skill published on SkillHub under the category of "ai-agent." Published by RealQ (Version 1.0.0), it adheres to the standard SKILL.md format. With a tag of `requires_api_key=false`, no separate API Key configuration is necessary for this Skill. This article objectively outlines its capability boundaries: the official description is empty, and page scraping failed. Since installation commands and configuration fields are not provided in the official documentation, unverified installation steps are intentionally omitted. Instead, only the official address and items requiring confirmation prior to integration are provided (specifically: verifying runtime support for SKILL.md; noting that with 1,473 downloads and 0 installations, the verification status remains uncertain; treating rating fields as unsuitable indicators of quality; and emphasizing that computer operation Skills must be utilized under controlled conditions within a sandbox).
Read MorePedagogical Explanation: An Agent Skill for Cognitive Psychology-Based Learning Strategy Recommendations
教学讲解 is a community-education skill published on SkillHub (version 2.0.2) by user_3c6cb52e. It belongs to the education category and requires no API key. The skill provides in-depth exploration of teaching explanation, grounding learning-method recommendations in cognitive psychology and emphasizing learning correctly rather than learning more. Its trigger word is tutor-explain. This article describes its positioning, core features, how to install and enable it, typical usage patterns, applicable scenarios, and notes/cautions.
Read MoreProcurement & Supplier Management | Guo Yiyi: Turning Supplier Evaluation into a Reusable Scorecard
This article introduces the SkillHub community skill 'Procurement Supplier Management | Guo Yiyi': it takes supplier quotations, delivery times, quality, payment terms, and fulfillment records as inputs, and produces supplier scorecards, tiered lists, and negotiation suggestions. The article verifies the publisher, version, category, the tag requires_api_key=false, usage boundaries, and discrepancies in popularity data. It explains that the skill is suitable for existing data evaluation scenarios for procurement managers, supply chain teams, and finance teams, and provides the official URL; since the materials do not include installation commands or reproducible examples, they are not fabricated.
Read Moredsh-logistics: Managing project logistics in DSH with modular expert review and optional repair
dsh-logistics is a DeepSeek Harness workflow plugin (v0.3.0, MIT) maintained by whitequeen306. Upon installation, it provides seven commands including /logi: it first performs deterministic scope collection, then has subagent experts write handoffs, and finally the main session model serves as the chief editor to output a module-aggregated report.md and verdict.json. The machine only validates the contract, without health scores or baseline gates. An optional /logi-fix command performs dry-run or whitelist-based fixes on direct items. Requires DSH 0.2.1+ and Node ^22.19.0||>=24.
Read Moredsh-breach: DSH Prompt Injection and Rollback-Capable Host Patches
dsh-breach is a DSH admin-security plugin maintained by zhuchongba-01, derived from dsh-purge. It provides prompt injection (including $DSH_HOME/prompt-inject.md and encrypted defaults) and rollback-capable string patches for the host source code; it settles 72 items on DSH 0.2.1-rc.5. It can only be installed from GitHub, with the Web surface command: dsh plugin --profile web add github:zhuchongba-01/dsh-breach. It includes happy_apply/revert, a net-scope request gate, /dsh-breach same-origin routing, and a floating panel. Please be aware of the risks associated with modifying the host source code before use; it is provided under the MIT license disclaimer and is intended for local learning and research only.
Read MoreWorkBuddy Expert Creation Skill: Compiling Business Background into Pasteable Expert Descriptions
This skill helps users create high-quality AI expert descriptions for Tencent WorkBuddy. Through a round-table style guided conversation, it collects your professional background, core use cases, and collaboration preferences step by step, and generates an expert description ready to paste directly into the WorkBuddy expert creation UI. According to the official brief, it is suitable for employees across enterprises, government/public institutions, education and research organizations, as well as individuals and freelancers. Its tag is requires_api_key=false, meaning no API key is required. It can be triggered when users mention WorkBuddy, want to create an AI expert, don’t know how to describe an AI assistant for their industry, want AI to better understand their business, or say “help me create an expert.” This article has verified the official URL, the SkillHub page coordinates, and the version. It notes that installation should follow the guidance on the SkillHub page. Since no standard installation command has been confirmed, no command is provided. Ratings, downloads, and installation data come from pre-fetched materials; it is recommended to reconfirm before citing.
Read MoreDSH-Chain: Making DSH Remember Decision Rationale
canglight/dsh-chain (dsh-chain-shared) is a memory plugin for DeepSeek Harness that records judgments, evidence, hesitations, and conclusions into local JSONL files. It features half-life-based forgetting across eight memory pools and injects content into the system prompt each turn within an `injectMaxChars` budget. Additionally, it supports jieba+IDF retrieval, `chain_about` metacognition, a 3D Nebula Tab, and nocturnal reports. This article provides installation and usage instructions verified against the SkillHub/GitHub README, covering link-based installation, npm dependencies, and considerations for session sharing.
Read MoreBatch Collection of Official Account Article Content: Introduction to wechat-article-search-api-skill
The wechat-article-search-api-skill is a WeChat article collection workflow based on the BrowserAct WeChat Article Search API template, published and maintained by @ccmagia2-gif on ClawHub, version v1.0.1, under the MIT-0 license. By inputting search keywords and optional publication date filters, users can obtain structured article data through a single Python script call, including original article links, publication dates, author or WeChat public account names, cover image URLs, full article body, and titles. It utilizes a pre-configured browser automation workflow to execute tasks, avoiding AI hallucinations, eliminating the need to handle CAPTCHAs and IP geolocation restrictions. Compared to purely AI-driven solutions, it executes faster at lower costs and includes built-in one-time automatic retries. The article introduces the Skill's positioning, core features, installation command (openclaw skills install @ccmagia2-gif/wechat-article-search-api-skill), BROWSERACT_API_KEY configuration process, input parameters, typical usage examples, execution status monitoring, output fields, error handling logic, as well as applicable scenarios including content monitoring, media research, trend tracking, knowledge base construction, and competitive analysis, along with usage limitations.
Read Moredsh-memewiki: A Built-in Meme Encyclopedia for DSH to Understand Group Chat Memes
dsh-memewiki is a DSH plugin maintained by thawline, featuring a built-in offline database of approximately 6,696 memes and slang terms ported from AstrBot's memewiki. It triggers a pre-check on messages that appear to be inquiring about memes; if a match is found, the specific entry is injected into the model. It also provides a `search_meme_wiki` tool and entry management via the settings page. To install, run `pnpm add github:thawline/dsh-memewiki` in the profile directory and add it to `dsh.profile.bundles`. Configuration is done via `~/.dsh/memewiki.json`. It requires Node.js ≥ 22.13 and is released under the MIT license.
Read MoreConversation Forking (Creating Branches): Fully replicate the entire working context of a long-running conversation into a separate branch.
Session Forking is a universal Agent Skill that comprehensively duplicates the workspace of a long conversation—including context, confirmed conclusions, completed steps, and tool results—into an independent branch. While the original branch remains intact, the new branch can diverge and re-execute from any specific node or explore multiple paths in parallel. This feature is ideal for scenarios sensitive to context continuity, such as dialogue, task execution, planning, coding, writing, and research. The forking object targets the workspace rather than the chat log, powered by a universal 'fork-core' engine for cross-product compatibility; no API Key is required for usage. This article outlines its positioning, core capabilities, trigger mechanisms, typical use cases, and applicable boundaries, while noting the absence of installation commands in the provided materials.
Read MoreDeconstructing Book Selling Points | Sun Sanye's Publishing Skill: Turning a Manuscript into Textually-Grounded Selling Points
"Book Selling Point Breakdown | Sun Sanye Publishing Skill" is an Agent Skill based on the universal SKILL.md format, maintained by community publisher user_6509230e. It is categorized as content-creation, version 1.1.1, sourced from the community. The skill generates a book marketing selling-point sheet from manuscripts, synopses, tables of contents, sample chapters, PDFs, Word files, screenshots, or pasted content. Each selling point is backed by source excerpts, includes page-number clues, and is mapped to target audience segments; star ratings are left blank to be filled in after user research. When needed, a Douyin remix video script can be generated. It is applicable to book breakdowns, selling-point extraction, target reader analysis, new-release marketing materials, and product-promotion video scripts. The article covers positioning, core features, official address and activation methods, typical use cases, and precautions.
Read MoreDouyin Content Engine: Integrating Topic Selection to Comment Operations into a Single Delivery
An introduction to the "Douyin Content Engine" by Lao Xie (Village Chief) on SkillHub: it compresses topic selection, spoken scripts, titles, cover copy, publishing strategy, and comment-section operations for Douyin spoken videos into a single full-pipeline delivery checklist. The wealth vertical receives targeted tuning covering hook preferences, posting-time patterns, and comment-section SOPs; finance-related content is subject to three outright prohibition triggers—income promises, recommending purchases on behalf of clients, and exaggerated returns—and delivery is allowed only when the automated check yields HARD=0. The core methodology includes a five-part spoken script (a 3-second hook, pain-point resonance, three actionable points, a memorable takeaway line, and conversion instructions, proportionally scaled by video length), five title formulas, and three seeded comment hooks. Sibling tools include the Viral Oracle (a pre-publish health check), the First 3 Seconds Hook Lab (for refining openings), and the Douyin Viral X-Ray (for dissecting benchmarks); this product focuses on production. The article also outlines trigger words, typical use cases, applicable scenarios, and limitations. Because the official page could not be retrieved this time, installation commands and configuration paths are not speculated.
Read MoreJiahao Model: DSH Persona Plugin — Change the Approach, Not the Delivery Standard
Jiahao Mode (dsh-jiahao-mode) is a fun character/costume plugin for DSH maintained by Rosynb666. It injects four behavioral guidelines: accepting jobs, engaging in combat, lagging, and gloating, with three adjustable levels of "toxicity." It supports soft mutual exclusion with other character plugins. The top-right corner provides a "Boldness" panel, kill notifications, and round settlement cards. The commands `/advance`, `/lagged`, `/niwan`, and `/career` allow users to progress, retry, and check battle records. Install via `dsh plugin --profile desktop add github:<owner>/dsh-jiahao-mode`. This is an unofficial fan-made work licensed under MIT and does not contain any game art assets.
Read MoreTersio: Token Savings and Usage Tracking for DSH
Tersio (@krtclcdy/tersio), maintained by KurutoDenzeru under the MIT license, is designed for users of token-billed coding agents. On the same extension tree, it provides Caveman (concise replies), RTK (compressed terminal output), Ponytail (slim diffs), optional pstack and Headroom, along with Combo presets and a local Dashboard. It supports MCP statistics reading and mode switching. DeepSeek Harness can be installed via `dsh plugin --profile web add npm:@krtclcdy/tersio` or `tersio install --host dsh`, while desktop users can add the npm package name through the application's plugin panel. Default configurations are stored in `~/.tersio/settings.json`. It is ideal for high-frequency users; it is recommended to review the source code and process permissions before installation.
Read MoreCo-creation Coach for Huaqiao Real Estate New Media Broadcasts: Turning Topics into Filmable 600–1000 Word Scripts
The "Huaqiao Real Estate New Media Spoken Script Coach" (Huaqiao Real Estate New Media Script Coach, @user_6414fc08/real-estate-script-coach, version 2.0.1, category: content-creation, source: community, tag: requires_api_key=false) bridges the gap between topic selection and filmable scripts for real estate short videos. It accepts two types of inputs: real estate topics (titles, one-liners, keywords, or scattered ideas) and verbal customer success stories. The final script is rigidly organized into four actions: "Hook First → Core Viewpoint → Illustrative Examples → Call to Action." It generates a 600–1000 word spoken script, requiring data verification and the inclusion of must-mention disadvantages before finalization. Upon completion, it outputs a ready-to-copy and editable .md document. Additionally, the script must adhere to the "512 Convention," with specific terms subject to SKILL.md. The article explains the tool's positioning, four core capabilities, and the verification boundaries for installation and activation (as the page does not provide installation commands or explanations regarding differences between Cursor / Codex CLI / Claude Code, specific details are not fixed here). It provides two directly reproducible input formats with expected outputs, and lists applicable scenarios along with five points for attention: external data must still be independently verified; storyboards are not produced; word count must be managed by the user; convention terms must be read first; and page ratings and download metrics may fluctuate (the current crawl shows a rating of 100,000, 2,136 downloads, and 0 installations, which differs from numbers in task clues; real-time page data should be referenced when citing).
Read Moredsh-live-compact: Compact on the Fly While Preserving Prefix Caching
dsh-live-compact is a DSH community plugin maintained by MerkyorLynn. When the context exceeds approximately 20,000 tokens, it performs an in-place spine cut on the just-completed turn without calling the model, retaining the prefix to maintain cache hits and delaying the official 80% major compression. It supports headless mode and web/Desktop preset replacement, though the TUI is currently unavailable. Installation requires the DSH 0.2.0-rc.2 line and Node 22.19+, with the current package version being 1.0.4. The cut original text is stored in `~/.dsh/compaction-spine/shadow`. The README includes evaluation data for 150 questions and 20 WebUniverse questions. MIT licensed.
Read MoreBull Stock King: Institutional & Hot Money Resonance Stock Selection – Consolidating Institutional Capital and Hot Money Trends into a Single Screening List
Bull Stock King - Institutional and Hot Money Resonance Stock Screener is an A-share stock screening Skill published on SkillHub (community source, version 1.0.0, category: professional). It first screens stocks using institutional capital, then screens based on hot money movements, takes the intersection of the two, and identifies individual stocks that attract attention from both types of funds. It provides three premium strategies: Mainline Leader Capture (for short-term hotspots), Swing Hidden Entry (for mid-term swing trading), and Small-Cap High-Quality Growth (for growth dark horses). The documentation states requires_api_key=false. The article introduces the dual screening logic, three strategies, verifiable deployment information, typical usage, applicable limitations, and truthfully states that the documentation does not provide installation commands, invocation examples, or internal parameters of the strategies.
Read Moredsh-auth-cookie-guard：Reserving Request Header Budget for Local Listeners and Retiring Old Startup Cookies
Introducing the DSH Community Host plugin **dsh-auth-cookie-guard** (jitianjiyan). Repeated launches with a fixed profile lead to the accumulation of `dsh-auth` startup cookies on 127.0.0.1, causing the Cookie header to exceed Node’s default 16 KiB limit and resulting in 431 errors. This manifests as an unrecoverable failure in desktop welcome authentication. The plugin increases `maxHeaderSize` (default 64 KiB) after the web server is ready, preserves a specified number of cookies (`keep`) and those for the current port in request order, and removes the rest in batches. Install via `dsh plugin add dsh-auth-cookie-guard`, with optional `maxHeaderBytes` and `keep` configurations. Includes notes on security boundaries and multiple instances on the same host.
Read MoreExam Paper Layout: A3 Two-Column Paper, Sealing Line, Auto-Statistics of Question Numbers and Scores, Answer Sheet Separation: Format Word Drafts into Formal Exam Papers Locally
Introducing the community Skill "Exam Paper Layout - A3 Two-Column - Sealed Line - Auto-Statistics - Answer Sheet Separation" (@user_9b6e0382/satloot-exam-paper, v0.1.0, community, category: education, rating: 100000, downloads: 1521). It accepts drafts written in a specific format (main section titles with total scores, numbered sub-questions, options A/B/C/D, and [Answer] markers) and formats them into A4 single-column or A3 landscape two-column layouts with a seal line on the left and writing fields on the front page. It automatically generates headers, exam time/full marks, question number/score tables, and page numbers. It renumbers questions, verifies score sums (alerting if totals mismatch), calculates points for blanks, and arranges multiple-choice options efficiently. Graphics, formulas, and tables are preserved (auto-scaling if too wide), while solution-based questions get blank space. It outputs a separate answer sheet and a question statistics table. Results are saved as new files without altering originals. It works offline (requesting permission to install python-docx if missing) and supports converting to PDF via Word or WPS. It does not contain built-in exam questions or formats mimicking standardized tests like the Zhongkao or Gaokao.
Read MoreOctopus | User Scenarios and Task Analysis: Turning Comments and Discussions into a Task List
**Introduction to the community-published Skill "Octopus | User Usage Scenario and Task Analysis"**, maintained by user_2d9d1ec9, version 1.0.0, categorized under data-analysis. It requires no API Key and is a top-level Skill manually triggered by customers. It takes Amazon Product Reviews, Octopus public discussions, and users' existing data as inputs, decomposes them along four dimensions—scenarios, goals, obstacles, and expected results—and outputs three categories of task lists: functional tasks, emotional tasks, and social tasks. It is suitable for JTBD (Jobs to be Done) user task analysis. This article describes its positioning, verified core capabilities, official entry point, typical usage, and limitations. The official installation command and configuration fields could not be verified at the time of scraping and were not fabricated.
Read MoreAgentCompare (dsh-agent-compare): Real execution of two versions for the same requirement; adopt after testing and review.
AgentCompare is a DSH community workflow plugin (MIT licensed, preview version 0.2.0-alpha.18) that generates two Vite frontend implementations in parallel for the same requirement, builds each one, and provides A/B preview in the conversation. It supports "two strategies with the same model" or "two models with the same requirement," binds them to unified acceptance criteria for blind selection, and after the reveal, reviews and writes back to the repository. This article, based on the official README, covers installation commands, interface operations, and applicable scope.
Read Moredsh-voice-mode: Full-duplex voice input and sentence-by-sentence playback within DSH sessions
dsh-voice-mode is a DSH client plugin maintained by qishuilalala that provides streaming local speech recognition into drafts during conversations, sentence-by-sentence reading with real-time subtitles, voice interruption, and optional wake words. ASR inference runs on the host side without requiring an additional API key, while TTS defaults to Edge cloud with optional VITS/Kokoro backends. Based on the official README, this article describes the installation command dsh plugin --profile web add dsh-voice-mode, typical operation gestures, settings entry points, and common configuration keys, while noting plugin permissions, online reading behavior, and compatibility boundaries.
Read MoreFacade Detailed Design Drawing Review and Connection Optimization Expert: Systematizing Drawing Review and Connection Optimization into a Reusable Process
This article introduces the SkillHub community skill "Curtain Wall Detailed Design Review and Connection/Node Optimization Assistant - 鹏哥撩工程 Ai" (v1.0.0, professional category, requires_api_key=false). Targeted at curtain wall designers, detailers, contractors, construction supervisors, and owner-side engineering managers, it consolidates eight capabilities into a fixed workflow: construction drawing review, curtain wall system analysis, connection/node detailing optimization, waterproofing and drainage analysis, embedded parts and connection system inspection, interface coordination with the main structure, construction quality control, and leakage troubleshooting. The article provides the official enablement entry point, recommended input fields, three types of reproducible examples (corner node optimization, post-rain leakage investigation, and as-built acceptance checklist), four output skeletons (drawing review comments, connection optimization plan, quality checklist, and leakage risk analysis report), as well as four guiding principles—node priority, waterproofing priority, constructability, and closed-loop management—and handling guidelines for issues such as sealant joint cracking, water leakage, and embedded part misalignment. It also clarifies the boundaries of the skill, stating that it does not replace structural calculations, professional testing, construction method approvals, or supervisory acceptance.
Read Moredsh-purge: Local Compatibility Layer for DeepSeek Harness Desktop
dsh-purge is a DSH client plugin (v1.3.6, MIT) maintained by 555bbbAAA based on upstream 1.1.71. Targeting the Windows desktop host (version 0.2.1-rc.3 segment), it provides patch-to-package swapping, runtime self-healing, tri-state updates, and CLI status queries. Based on README and INSTALL.md, this article describes the compatibility scope, core capabilities, PowerShell installation and registration steps, verification methods (--status / --swap-asar / --uninstall), as well as local file modification permissions and known limitations.
Read MoreOverseas Litigation Case Database: Cross-Jurisdictional Precedent Search and Compliance Insights
Overseas Litigation Case Database (@user_44ebfc99/cue-overseas-litigation, v1.0.5, community/professional) is a retrieval-based Agent Skill. It centers on a specific topic to retrieve public precedents and regulatory announcements from major legal jurisdictions, summarizes the structure of causes of action, judicial tendencies, and actionable compliance insights for Chinese entities. Core requirements mandate that key assertions include original citations or links, and that no precedents, case numbers, penalty amounts, or legal provisions be fabricated. In cases of insufficient sources, the system must state this explicitly and substitute with verifiable official documents. For non-English jurisdictions, original terminology is preserved with supplementary Chinese explanations. Materials are organized into three layers: scripts/ provides scripts for case database retrieval, judgment extraction, and trend aggregation, while references/ offers manuals on legal jurisdictions, court official websites, and citation formats. Running the skill requires an API key and supports three installation methods: Cursor, Codex CLI, and Claude Code. It is suitable for preliminary research on cross-border compliance regarding sanctions and export controls, but does not constitute legal advice.
Read More