Short Drama Scriptwriting: Assembling Any Material into a Shootable Screenplay
Introducing the community Skill "Short Drama Script Writing" (@indiv-xunaji/wwdj). It registers all materials provided by users as registry assets, using JSON-linked data as the primary data source to assemble story structure, storyboards, dialogue/narration, and per-shot prompts layer by layer, ultimately rendering a complete version ready for direct copy and use. Characters, scenes, and shots are linked by ID, and changes are propagated and synced in batches by index. Storyboards and prompts follow platform-agnostic general film/video and image generation standards; the aspect ratio must be clarified before use, and image quality is determined by the generation tool. In addition to the complete workflow, it can also independently perform novel-to-episode adaptation, scriptwriting, dialogue, sound effects, image generation prompts, video generation prompts, motion/animation effects, special/magic effects, scene staging direction, fight scene storyboard design, continuation writing, continuity review, and delivery/external integration. Version 8.3.10, category content-creation, no API Key required. Page rating 100000, downloads 854, installs 0. The official page was inaccessible at the time of writing, so no installation command is listed; it is recommended to refer to the page or client installation entry point.
Read MoreExpressiveness | Eloquence and Epigrams: A Systematic Expression Training Program
Expression Skill | Articulate Eloquence and Piercing Words is an expression training skill designed for the generic SKILL.md format, produced by Cai Fu Academy, maintained by Village Head Lao Xie, and categorized under 'community' and 'content-creation' (Version 1.0.6). It begins with a four-layer diagnosis (fear of speaking, unclear articulation, shallow depth, and lack of persuasion) followed by targeted training modules covering structure, visualization, rhythm, refinement, punchlines, sense of prestige, vocal quality, and stage presence. The package includes scripts for five scenarios—reports, meeting contributions, impromptu speeches, roadshows, and event hosting—along with a 21-day daily training schedule and nine 'on-site emergency' protocols. It also offers dialog practice, video self-assessment, one-click image generation, and a fill-in-the-blank template for the "Expression Training Plan." This tool is suitable for individuals preparing for performance reviews, job reports, roadshows, hosting gigs, or seeking to improve their public speaking. However, it is not intended for those with severe social anxiety or expression fears stemming from psychological trauma, nor does it facilitate the fabrication of false personal experiences.
Read Moredsh-background: Restore full-window background images for DSH Web/Desktop
DSH 0.2.x removed the built-in background image configuration. The community plugin dsh-background provides image selection, restoration to default, and removal via bundles in the bottom-right corner of the Web/Desktop interface. Images and settings are persisted in $DSH_HOME/dsh-background/, and automatic transparency is applied to regions like the sidebar. Based on the official README, this article describes its features, installation steps (cloning into the plugins directory, linking profiles and registering bundles, and performing a full restart), as well as image limitations and loopback security boundaries.
Read Moredsh-channel-hub: Remotely Driving Local DSH via Feishu Long Connection
dsh-channel-hub (Warrenz23, MIT, v0.2.0) turns Feishu into a DSH remote console: the client initiates persistent WebSocket connections to the server, so no public IP is required. It provides a 127.0.0.1 graphical configuration panel, whitelist support, progress cards and approval cards, and supports images, voice messages, files, and chat commands. Based on the GitHub README, this article outlines installation (dsh plugin add / node bin/install.mjs), the 3-step console workflow, and configuration essentials, and explains permissions and security boundaries.
Read MoreIntelligent Posture Analysis Skill: Upload a standing photo for segmented posture assessment and improvement recommendations.
The "Intelligent Posture Analysis Skill - Your Personal Body Posture Assessment Expert" is maintained by community publisher user_7ff8503b. Classified under life-service and sourced from the community, it is tagged with `requires_api_key=false`, meaning no additional keys are required. The Skill accepts standing photos from the front, side, back, half-body, or full-body views. It triggers when users request posture analysis, posture checking, or body posture assessment, with official trigger keywords including "posture analysis," "posture assessment," "check my posture," and "check my standing posture." Its capabilities cover analysis across six major regions: the head and cervical spine, shoulders, thoracic spine and trunk, pelvis, knees, and feet. It also features problem identification with severity ratings, chain-link correlation analysis, targeted exercise recommendations, daily habit advice, and HTML visualization reports. Technically, based on a professional posture assessment knowledge base, it adopts conservative evaluation principles to avoid over-diagnosis, includes a built-in medical reminder mechanism, and carries a disclaimer. The article outlines trigger methods, reproducible analysis workflows, and applicable scenarios such as self-examination for office workers, pre-workout assessments, progress tracking for yoga and Pilates, and physical rehabilitation references. It emphasizes that the tool serves as an auxiliary reference and cannot replace professional medical diagnosis, while noting that official documentation does not provide installation commands or details regarding variations across different clients.
Read MoreQuant Investment Cheat Code Pro: Backtesting Strategies with a Single Prompt to the Agent
# Foreword Transforming a stock-picking idea into a verifiable backtest result typically requires building data pipelines, writing factor calculations, running simulations, and manually calculating returns and drawdowns while staring at charts. Many people get stuck at the very first step: they cannot code, or perhaps they can, but lack the time for maintenance. The philosophy behind this Skill is to compress this entire workflow into a single conversation—you state your objective, and it handles strategy execution, report generation, and risk alerts. Described below is a quantitative strategy Skill designed for Large Language Model (LLM) Agents. It originates from the community rather than an official vendor, and its capabilities should be interpreted based on the page description. # What is this **One-sentence positioning:** A Skill named "Quant Investment Cheat Sheet pro" that enables Agents to perform strategy backtesting and market state judgment using natural language, located at @user_3af33138/qua...
Read Moredsh-model-router-lite: Model Selection Suggestions for Sub-agents Based on Task Type
Model Router Lite is a DSH Host plugin maintained by WangZW928. It uses three types of task preferences—routine/complex/review—plus reusable strategy prompts to guide the parent agent to check list_subagent_models and explicitly choose a model before creating a native sub-agent. It does not grant automatic authorization and does not add any routing-model calls. Requires DSH 0.2.1-rc.5 or above and Node 22 or above. After installing dsh-model-router-lite@0.1.0, you must allow the routing in DSH, configure routes, and it takes effect for newly started top-level conversations. It is an advisory strategy and not a forced routing.
Read MoreExposing DSH-Configured Models as OpenAI-Compatible HTTP Endpoints
dsh-endpoint is a DSH Cordis plugin maintained by wangzhongyang. It provides OpenAI-compatible /v1/chat/completions and other endpoints within the host process. Requests are forwarded via ctx.llm.stream(), and no vendor credentials are stored separately. It supports model listing, streaming chat, optional session mirroring, and model aliases. Installation requires DSH Desktop web profile 0.2.1-rc.x, using the command dsh plugin --profile web add dsh-endpoint. By default, it has no authentication or rate limiting, so it should be bound to 127.0.0.1. MIT 1.1.0.
Read MoreXZZ Short Video Batch Automated Editing: Input Copy to Generate Vertical Short Videos in Batch
Batch Short Video Mixing (also named XZZ Short Video Factory on the SkillHub page) is a community Skill published by user_9404a307, version 1.0.3, and categorized under design-media. By inputting a text script or a reference video, it automatically batch generates vertical short videos featuring covers, subtitles, voiceovers, and BGM. It features built-in Piper offline TTS for free voiceovers, PIL-rendered first-frame title pages as covers, ASS subtitles with emphasis highlighting and text wrapping for long sentences, and a task fingerprinting system for cumulative deduplication. Outputs are saved to date-named desktop folders, with dates and duplication rates included in the file naming. With a zero-threshold trigger, simply saying "Help me make 5 videos" instantly produces the videos. No installation command is provided on the official page, and the skill carries a `requires_api_key=true` tag, so the key type and configuration location must be confirmed before enabling it. It is applicable to batch mixing scenarios such as travel promotion, dining exploration, real estate marketing, and knowledge sharing.
Read Moredsh-session-delete: Permanently delete sessions in DSH
By default, DSH only archives sessions; the JSONL files on disk, projection cache, workspace index, and plugin sidecar data are not cleared automatically. dsh-session-delete (wangyueyueye, admin-security) scans and displays all data locations before deletion, offering three scope levels and normal/overwrite erasing. It executes the deletion via the owner API and path gate, writing an audit JSONL. Based on the official README, this article describes installation methods (local/Git/npm), menu interactions, and known limitations (including that logs for already-uploaded models cannot be retracted locally), and reminds users that Host plugins run in-process, so the source code should be reviewed before installation.
Read MorePrompt Revision_Human-Like Tone: Translate "No AI Tone" into a Reusable Prompt Framework
Prompt Modification — “Human-Like Feel” is an AI-agent Skill on the SkillHub community, hosted at @indiv-haikousohot/zhenrengan-prompt, version 1.0.0, requires_api_key=false, with 100,000 ratings, 847 downloads, and 0 installs. It turns colloquial instructions such as “make it feel human” and “avoid an AI-like feel” into structured prompts. Its approach is derived from the L3 material micro-level in a five-layer realism architecture, covering three components: a 13-field schema, emotion inference, and photography parameters. It triggers when a user asks for a realistic, human-like feel. This article verifies the official introduction against the statistics, clarifies its positioning, capability boundaries, and usage steps, and notes that the materials do not provide an installation command, a field list, or reproducible examples, so none are fabricated. A typical workflow is described in four steps: trigger detection, preserving the user’s original request, checking field by field, and filling in the results accordingly. It also states that it applies to realistic portraits, does not apply to non-realistic styles, and lists three common pitfalls.
Read MoreLivestream Compliance Violation Check | 郭亿易: Instant Output of Risky Script Sentences and Suggested Rewrites
Introduces the community Skill "Livestream Prohibited Word Review | Guo Yiyi": input livestream scripts, platform rules and product qualifications to generate a risk sentence/phrase list with risk levels and compliant rewrites, suitable for livestream compliance and content review experts. The Skill follows the standard SKILL.md format, is categorized as 'professional', has a version number of 1.0.0, and is API-key-free. This post explains the input preparation methods, verifiable prerequisites for activation, input organization for API calls, applicable scenarios and limitations, and notes that installation commands and usage examples should be primarily subject to this respective official website.
Read MoreAgent Autonomous Loop Engine: Self-Correction Closed Loop After Errors
Agent Autonomous Loop Engine is a community Skill that provides AI Agents with runtime self-evolution capabilities. Located at @user_638e256a/ai-loop-engineering, it is version 1.0.0 and categorized under ai-agent. It automatically triggers a perception–evaluation–evolution–verification closed loop when an Agent makes mistakes, is corrected by users, exhibits expectation deviations, or shows performance degradation. It includes error capture, root cause diagnosis, policy generation (patch/refactor/extend/deprecate), hot updates, and meta-evolution introspection. This article outlines its positioning, verified core features, usage, and applicable scenarios, and candidly notes that installation commands and official examples are missing due to the page being unreachable; users should verify the official page before use.
Read Moredsh-model-toggle: Hide and Restore by Model in DSH Settings Page
dsh-model-toggle is a DSH model inference plugin maintained by wangnian886. It provides a "Model Toggle" panel at the bottom of the Settings → Models page, allowing users to hide or display each model for custom llm-pi-ai providers with a single click. Once hidden, the model disappears from the bottom selector in the conversation, with its entry archived in dsh-model-toggle.archive.json within the profile directory and ready to be restored at any time. Requires DSH 0.2, and supports installation via GitHub source or Release tgz. The plugin prevents hiding the last visible model and the current default model, and automatically restores settings upon uninstallation or disabling. It is ideal for users with numerous models who need to streamline the selector without manually editing YAML files.
Read Moredsh-scheduled-send: Schedule sending user messages in the DSH input box
dsh-scheduled-send is a DSH client plugin maintained by wangjiaxi (v0.2.2, MIT). It turns the send button in the Web input box into a Feishu-style split capsule: click ▾ to choose a time, and the host writes a draft into $DSH_HOME/storages/dsh-scheduled-send/tasks.json. When that time arrives, it is delivered as a normal user message via sessionController.prompt, which is different from schedule_* reminder frames. It still executes according to the epoch time even if the page is closed, the session is switched, or the host is restarted; it supports presets, list-based cancel, and an HTTP API. Install: dsh plugin --profile web add github:wangjiaxi/dsh-scheduled-send; it takes effect after refreshing the page. Text only; depends on webServer and sessionController.
Read Moredsh-peak-valley：Display peak/valley countdown below DSH input box
dsh-peak-valley is a DSH client plugin maintained by wangjiaxi. It displays Beijing Time peak/off-peak hours and the countdown to the next switch to the right of the web input box dock and the context indicator. Peak hours are 9–12 and 14–18 on weekdays; all other times, including weekends and public holidays, are off-peak. The holiday calendar is stored in client.js and covers 2026–2027. It is a purely client-side plugin, released under the MIT license. To install: dsh plugin --profile web add github:wangjiaxi/dsh-peak-valley.
Read Moredsh-wsl-sidebar: Running Linux-side Agents with WSL in DSH
dsh-wsl-sidebar is a client plugin for DeepSeek Harness that runs read, write, and search operations inside a specified Linux distribution via wsl.exe, avoiding write failures and ACL issues via \\wsl.localhost. It provides three WSL tabs in Settings, an entry point to a WSL directory for creating new sessions, and six tools: wslx-run, wslx-read, wslx-write, wslx-edit, wslx-search, and wslx-bg. To install, run dsh plugin add dsh-wsl-sidebar or github:Wang1462604987/dsh-wsl-sidebar, and restart the Harness after installation. Released under the MIT License. Maintainer: Wang1462604987.
Read MoreAuto-Generating a 100-Article Series of Diagnostic Notes: Batch-Producing Systematic Content from “Domain + Number + Disease Name”
This is a community skill on SkillHub located at @user_84435c73/runge002, version 1.0.1, categorized under content-creation, sourced from the community, and requires no API key. It is designed to accept inputs in the format of "domain + number + diagnosis name" and automatically generate a series of diagnostic notes for any domain. Its official description claims coverage of 100 diagnoses across computers, workplace, brands, marriage, cars, education, health, and gaming. Core capabilities include swapping domains, filling numbers up to 100 entries, unifying a symptom–cause–remedy structure, and cross-referencing entries to create a coherent series. The documentation does not provide a standard installation command or directory structure, so the article only lists the official URL, page, coordinates, version, and source, with activation instructions subject to the page's guidance. Usage follows the description by extracting three input parameters, and suggests batch requests for 001–020 to avoid truncation from overly long single outputs. Key points include the number as the critical input for batch generation, the problem-oriented style of diagnostic writing, the need to verify cross-domain judgments independently, unspecified default formatting, and the fact that the page's rating of 100000 with 864 downloads and 0 installs indicates a small sample size and should be refreshed before citation.
Read MorePro Tool for Intelligent Contract Review and Key Clause Extraction: Breaking Down Contract Review into Actionable Checklist Items
This is a pay-per-use Skill costing ¥2, located at @user_6850f0da/pdf-contract-review-pro, version 3.0.6. Released by community developer user_6850f0da, it is categorized under "professional." It conducts a structured review of the entire contract, outputting clause summaries, numerical and logical verification, P0/P1/P2 risk grading, a 0-100 health score, direct replacement suggestions for clause modifications, negotiation priorities and scripts, and a checklist of missing clauses. It also calculates payment and performance deadlines based on statutory holidays and make-up workdays, explaining the number of extension days and the reasons. Payment is based on the A2M Smart Receipt Protocol (HTTP 402), with pay-per-use collection handled by Alipay AI. This article explains the core functions, prerequisites for enabling (requires_api_key, pricing_type=alipay), and applicable scenarios, without fabricating installation commands or invocation examples not officially provided.
Read MoreComic Creation Tool | Character Three-View Generator: From Web Novel Scripts to 3D Chinese Comic Character Design Sheets
This Skill solidifies the workflow of "reading synopsis → extracting core character traits → generating a standard three-view setting sheet." By inputting synopses from time-travel, historical, or fantasy novels and web fiction, it first parses the name, gender, precise age, and identity/occupation for each character, ensuring that major supporting roles are not omitted. It then generates a strict 16:9 horizontal character setting sheet for each character: with a pure white background, 3D Chinese animation style, a facial close-up on the left, and front/side/back full-body views on the right, featuring a three-layered clothing design. Duplicate outfits across characters are strictly prohibited. The article provides trigger words, a step-by-step reproduction process, and an image verification checklist. It also explains how to handle missing data: since the official page only provides the address, coordinates, and version 1.0.2—without specific command-line installation snippets or configuration fields—no unverified commands are included, requiring verification on the page itself. The parameter `requires_api_key=false` indicates that no additional API Key is needed. The article further outlines applicable scenarios and points of caution: the Skill cannot fill in missing ages or identities if they are not written in the source text, and the "three-layered design" and "no duplicate outfits" are style constraints rather than automatically guaranteed features. Given that the platform displays 840 downloads, 0 installations, and a community source, it is advisable to conduct a trial run before integration.
Read MoreMultimedia Generation Routing Orchestration: Downstream Routing and Confirmation Gates for I15 Intent
This article introduces the "Multimedia Generation Route Orchestration" skill on SkillHub (@user_989eb8f0/tri-mm, version 1.6.0, community source, category: ai-agent, rating: 100000, downloads: 855, requires_api_key=false). It processes multimedia generation intent I15 from Section III of the tri-intent snapshot; after identifying the broad media category, it dispatches tasks to the tri-image, tri-audio, tri-video, or tri-ppt sub-skills, while music creation is separately delegated to tri-music. Each sub-skill first generates a multi-dimensional design proposal for user confirmation before proceeding with generation. The skill activates when the downstream routing suggestion in the upstream snapshot points to it. It supports independent installation and contains two-state logic: a standard mode with upstream dependency detection and a guided installation mode. Due to a failure in scraping the official page, the final draft did not include unverified installation commands.
Read Moredsh-work-review: A DSH workflow plugin that generates ISO weekly reports from weekly journals and Dream.
dsh-work-review is a DeepSeek Harness workflow plugin (MIT, v0.5.0) maintained by Wang-Chuming. It reads Dreams, journals, ingests, tasks, and git logs within the current ISO week from an Obsidian-style vault. It writes data to `life/routines/<YYYY>-W<ww>.md` via CLI and displays this content in a read-only "Work Review" panel on the left side of the Web GUI, organized by day and week. Installation is performed via `dsh plugin --profile <profile> add github:Wang-Chuming/dsh-work-review`; it requires configuring `vaultRoot` and fully restarting the Harness. The package includes examples for nightly Dreams, optional review executions, and automation schedules. Both the panel and the GET API are read-only. This tool is suitable for developers who use vaults for daily notes and require weekly reviews within DSH; please review the source code and permission boundaries before use.
Read MoreHumanizing AI Output: A Skill for Natural Chinese Rewriting
Introducing the “AI-Tone Rewriter” skill on SkillHub, available at @kunlungrowth/ai-tone-rewriter. Published by Kunlun Growth, version 1.1.0, categorized under content-creation, enterprise-sourced, with a rating of 100000 and 830 downloads. It aims to refine AI-generated Chinese text into more natural, human-like expressions by eliminating clichéd phrases and mechanical sentence structures. The tag requires_api_key=false indicates no additional API key is needed. This article explains what the skill is, its core capabilities, installation access, and location. It also notes gaps in the verified materials—including no concrete installation commands, directory paths, configuration fields, differences across Cursor, Codex CLI, or Claude Code clients,or reproducible official usage examples—and therefore avoids speculation, recommending reliance on the official page. Use cases and caveats are also provided: it rewrites text without verifying facts, and human review is still required for official documents;the ability to adjust the rewriting intensity is not documented. The official page can be accessed on SkillHub.
Read Moredsh-obsidian-file-opener: One-click to open vault notes with Obsidian in the Harness Web GUI
dsh-obsidian-file-opener is a DSH client bundle plugin maintained by Wang-Chuming. It intercepts ordinary clicks on file references within the vault in the Web GUI: after the Host validates the path, the file is opened in Obsidian via the obsidian:// protocol; Cmd/Ctrl+Click still uses the built-in preview. It supports openers that hand off files such as .tex to VS Code, and provides the open_vault_file model tool. Installation is done with `dsh plugin --profile add github:Wang-Chuming/dsh-obsidian-file-opener`, with configuration written to `~/.dsh/obsidian-file-opener.config.json`; the Harness must be restarted for changes to take effect. Released under the MIT license, it invokes system openers with the permissions of the current DSH process.
Read Moredsh-plugin-scratchpad: Add Local Scratchpad to DSH Left Sidebar
The dsh-plugin-scratchpad, a DSH bundle plugin maintained by wake2008dingtalk-sys, adds a Scratchpad entry to the left sidebar; the main panel offers multi-line notes, a character counter, and a clear action, with content persisted locally via localStorage and not uploaded, and supports Chinese and English locales and DSH theme tokens. Install with plugin_manager install_bundle pointing to the absolute repository path; uninstall via remove_bundle @local/sidebar-scratchpad. Validated under Harness 0.2.1-rc.5. MIT licensed.
Read More