dsh-cookie-guard: Limiting Desktop dsh-auth Cookie Lifespan to Prevent Startup Crashes
The DSH desktop client is prone to startup failures due to the accumulation of dsh-auth-* cookies that remain uncleaned for 30 days combined with random port assignments, causing the Cookie request header to exceed Node's 16KB limit and triggering HTTP 431 or "Web authentication failed" errors. The community plugin dsh-cookie-guard, maintained by kevindawn11-hash, addresses this by setting cookieMaxAgeDays to 7 through a bundle patch in the desktop profile with no runtime intrusion, and provides prune-cookies.mjs to clean up stuck cookie stores. Installation requires Node >= 22 and DSH 0.2.1-rc.3+, and can be installed via the recommended command: dsh plugin --profile desktop add dsh-cookie-guard.
Read MorePlugin Workshop (dsh-plugin-studio): Classify, understand, clean, and package DSH plugins and skills on a single page.
dsh-plugin-studio (Plugin Studio) is a DSH client plugin maintained by KERBALMans that lets you manage installed plugins and skills within a browser single‑page interface. It provides interpretable three‑level automatic categorization, plain‑language AI explanations, separate last‑used timestamps for tools and calls, a new “Recently Run” filter added in v1.0.8, batch delete with restart prompts, and one‑click packaging. To install, enable it with one click via Settings → Plugin Studio, or install the 1.0.8 tgz from the GitHub Release. Requires Node ≥ 20 and DSH ≥ 0.2.0-rc.1. MIT licensed.
Read MoreAI Video Editing and Multimedia Content Generation (AIPDD): Delegating the Raw-to-Final Cut Workflow to Agents
AIPDD is an Agent-oriented video editing and multimedia content generation Skill, maintained by AIPDD and hosted at @org-bzwfrdvh/aipdd-video-clip. Version 1.0.2, sourced from enterprise, categorized under design-media, with the tag requires_api_key=true. It consolidates material processing, automated video assembly, short-drama and short-video production, as well as multi-model generation capabilities—such as text-to-image, image-to-video, and speech synthesis—into a single natural-language trigger. It ships with the AIPDD model gateway, compute configuration, and API Key provisioning. The article explains the problems it solves, the capabilities it covers, how to enable and configure it, and the considerations to keep in mind when integrating it.
Read MoreClinical Biostatistics: Unifying Two Common Hypothesis Tests into a Callable Skill
Clinical biostatistics is a data-analysis skill published on the Community, version 1.4.19, located at @user_f4fc8899/bio-clinical-biostatistics-integrated. It offers two common hypothesis tests: the Chi-square test (contingency tables) and the Welch's t-test (two groups of continuous variables), returning the statistic, degrees of freedom, and p-value. It operates as a registered hosted executor: accepting documented JSON input, returning structured content, and providing Markdown output when parameters are supplied. It does not execute upstream local toolchains or operate third-party accounts; deliverables are based on server-validated responses. Fetched metadata indicates requires_api_key=false, pricing_type=alipay, source=community, current page rating=100000, downloads=1135, and installs=0. Since the documentation lacks installation commands, directory paths, and configuration fields, the differences in integration with Cursor, Codex CLI, and Claude Code are not speculated upon; it is recommended to consult the official page for instructions. It is applicable to projects with existing dataframes requiring rapid completion of routine hypothesis tests and unified statistical standards. Limitations include coverage of only two types of tests, and the hosted execution model makes it unsuitable for scenarios requiring data locality (data not leaving the domain) or offline computation.
Read Moredsh_qzone_sticker: Enabling DeepSeek Harness to Silently Log into QQ QZone Locally and Concurrently Post Messages
The kdcadmin DSH networking tool plugin dsh-qzone (0.4.1) logs into QQ Zone via headless Chrome on the local machine, stores the session in ~/.dsh/, and provides tools such as qzone_status, qzone_recent, and qzone_publish. It supports image attachment, a daily limit, and a 60-second interval between posts, with a Web UI fallback when the album fails. Requires Node 22+ and a local Chrome installation. Currently, installation is primarily via a link inside the profile, and the settings entry is labeled “Post a ShiSay.” It is open source under MIT. Please review the source code and process permissions before use.
Read MoreShort Story Writing: Structure and Character in Medium-Length Narratives
This article introduces the "Short Story Writing" Skill published by user_3c6cb52e on SkillHub. It targets medium-length narratives and organizes the writing workflow around complete story structure, character development, and thematic expression. Crawled data shows the version is 2.0.1, categorized as content-creation, sourced from community, tagged requires_api_key=false, and the page snapshot displays a rating of 100000, 1198 downloads, and 0 installs. The available materials do not provide installation commands, differences between Cursor/Codex CLI/Claude Code, or official reproducible examples. Therefore, this article does not list unverified commands and usage, but only describes confirmed capabilities, activation boundaries, and trial suggestions.
Read Moreworkbuddy2everything: Transforming a Local WorkBuddy Gateway into a Multi-Client Shared Model Backend
`kangjiazhen/workbuddy2everything` (npm: `dsh-workbuddy-codex`) is a DeepSeek Harness model reasoning plugin. Upon loading, it detects and launches the wb2api gateway (7863), the built-in Responses bridge (7864), and the Anthropic Messages bridge (7865), and mounts the workbuddy provider via `dsh-llm-workbuddy`. It resolves the issue where requests fail despite the model list displaying correctly, caused by ports not starting after a restart, allowing Codex CLI, Claude Code, and the Claude desktop 3p to share the same gateway and the `api_key` from `config.json`. The installation command is `dsh plugin add git+https://github.com/kangjiazhen/workbuddy2everything.git`. Before use, it should be noted that the plugin spawns child processes with the permissions of the DSH process, and the source code and MIT license should be verified independently.
Read MoreHome Renovation Assistant: Comprehensive Pitfall Avoidance & Inspection Guide - Pengge on Engineering AI
This article introduces the community skill “家庭装修全流程避坑与验收助手 - 鹏哥撩工程 Ai” published on SkillHub. It covers the entire renovation journey—home handover and inspection, plan design, contracts and quotations, demolition and alterations, electrical and plumbing, waterproofing and masonry, carpentry and ceilings, painting, custom installations, final acceptance, and eco-friendly move-in—and provides twelve capabilities: renovation planning, raw-house inspection, needs planning, quotation and contract review, whole-house electrical/plumbing point planning, pitfall avoidance in demolition/alteration, electrical and plumbing acceptance, waterproofing and masonry acceptance, carpentry/ceiling and custom work acceptance, painting acceptance, main-material installation acceptance, and final snagging and move-in risk checks. The article describes its applicable scenarios, recommended input formats, official examples, output deliverables, and clarifies that it does not replace designers, home inspectors, licensed electricians, gas professionals, waterproofing installers, or testing agencies. No installation commands are provided in the official materials, and the tag requires_api_key=false.
Read Moredsh-figmark: Circling research figures in the DSH right-side preview and delegating annotations to an agent to modify the figure source
dsh-figmark is a DSH client plugin maintained by KafukaTree that enables freehand markup and annotations directly on PNG/SVG/PDF previews and local loopback web pages in the right sidebar. It writes sidecar annot.json files and synthesized attachments without modifying the original files. When sending, it composites coordinates, SVG/HTML element anchors, and change instructions into the user message, guiding agents to update the generation sources (e.g., Python, DOT, SVG). It supports installation via plugin marketplaces, npm, and local paths. Note that static HTML pages are read-only and cannot be annotated, PDFs have native annotation limitations, and the plugin runs with the permissions of the dsh process.
Read MoreAdobe InDesign 2026: Multi-Document Batch Layout and Script Automation
adobe-indesign-2026 is a skill designed for Adobe InDesign 2026, identified by the coordinate @org-k6bcf2fv/adobe-indesign-2026. It is version 1.0.0, published by "skill训练营," categorized under "design-media," and originates from an "enterprise" source. Its official description states that it generates and deploys InDesign 2026 scripts to achieve batch layout across multiple documents and workflow automation. The retrieved materials do not provide installation commands, configuration fields, or official usage examples. The tag requires_api_key=false indicates that no API key is needed. The text outlines the skill's positioning, verified information, and inferred usage methods, and recommends checking the official page before installation.
Read Moredsh-superagent: Proactive Coding Agent on Monorepo GitHub
dsh-superagent is a DSH bundle maintained by jwangkun, forked from deepseek-ai/dsh-superagent-prototype. It monitors issues/comments (with optional Feishu support) for a single GitHub repository, converts requests into tickets under Policy authorization, executes them in isolated sessions, and responds via PRs or other means. This article, based on the README, outlines installation commands, Superagent panel configuration, shared-account mode, and fork environment isolation. It emphasizes the project's positioning as a research prototype and the necessity of reviewing the source code and permissions before installation.
Read MoreFFmpeg Video Processing Pipeline: Solidifying Filtergraphs into a Pipeline
FFmpeg Video Processing Pipeline is a Skill published on SkillHub Community Edition, identified by the coordinates @user_3c6cb52e/ffmpeg-video-processing-pipeline. It is categorized under design-media with version 1.0.0. This Skill encapsulates the process of building filtergraph chains required for batch video transcoding, thumbnail sprite generation, and HLS adaptive bitrate packaging. Callers are only required to provide the input, output, and target specifications, while the Skill handles the construction and execution of command combinations. Verified capabilities include batch transcoding, sprite sheet generation, HLS packaging, NVIDIA NVENC hardware acceleration, and HDR tone mapping. It is tagged with `requires_api_key=false`, indicating no external API key is needed. Since the official page does not provide structured installation commands, this article only provides the SkillHub page and API addresses, and outlines the local ffmpeg and driver prerequisites required for NVENC. As the documentation lacks reproducible invocation examples and parameter structures, this article describes the workflow-level usage across the three stages of batch transcoding, sprite generation, and HLS, while highlighting potential pitfalls such as filtergraph errors, missing HDR metadata, NVENC concurrency limits, and path mismatches.
Read MoreDelegating Rejection Reasons to the Model in DSH Permission Approval
DSH native permission approval is binary—allow or deny—leaving the model unaware of the rejection reason. dsh-why-reject-me adds an optional "Rejection Reason" input field to the Desktop approval panel. If a rejection occurs with a non-empty reason, the text is appended to the tool's result, making it available for the model to read in the same turn; rejecting with an empty field behaves identically to the native function. Maintained by Jorgen233, this plugin (npm package: dsh-why-reject-me, v0.3.3) requires installation via DSH Desktop and a restart to enable. Reasons are recorded in the session history; this feature is not applicable to sub-agent sessions.
Read MoreChemical Heat Exchanger Design Calculation and Selection: A Process Flow from Selection to Mechanical Design
This article introduces the positioning and capabilities of the community Skill “Chemical Heat Exchanger Design Calculation and Selection.” The Skill identifier is @user_a8c636dd/heat-exchanger-design, version 1.0.0, source: community, category: professional, rating: 100000, downloads: 1215, tag: requires_api_key=false. It covers shell-and-tube, plate, double-pipe, spiral plate, and finned-tube heat exchangers. Capabilities include type selection; heat-transfer calculations (heat duty, LMTD, FT correction factor, inside/outside heat-transfer coefficients, overall heat-transfer coefficient, and required area); pressure-drop calculations (tube-side straight-run plus U-bends and inlet/outlet nozzles; shell-side using the Kern method); TEMA selection (combinations such as AES/BES/AET/CES); and mechanical design essentials such as tubesheet thickness, shell diameter, tube layout, baffles, materials, and expansion joints. The article notes that the official source does not provide a unified installation command; it should be enabled by referencing the identifier, and typical question sequences as well as applicability limitations are outlined.
Read MoreSafety Technology Skill: Integrated Technical Q&A on Fire Protection, Industrial Safety, and Special Equipment
This article introduces the community Skill "Safety Technology" ( @user_b28db9b4/anquanjishu , v1.0.0 , professional ) on SkillHub. It integrates three professional domains—fire protection technology, safety technology, and special equipment technology—to provide regulatory interpretation, facility configuration consulting, emergency response guidance, safety management improvement plans, and full lifecycle technical support for special equipment. The Skill is keyword-triggered and applies `requires_api_key=false`. The article compiles the five categories of capabilities and all trigger keywords listed in the official description, presents four organized typical dialogue scenarios, and notes limitations such as installation and activation being subject to the official page, the knowledge cutoff and specification versions not being disclosed in the provided materials, and a zero installation count.
Read MoreZGS-PPT: Career Planning Competition Reading-Style Roadshow PPT Generation DSH Plugin
Input Information: John-20260101's DSH workflow plugin zgs-ppt, designed for the 4th National College Students Career Planning Competition (Career/Job-seeking presentation PPT). It features 17 page types, 15–20 page reading-optimized layouts, and aligns with the structure of previous Gold Award winners. Using outline.json as the source of truth, it confirms the outline before rendering the pptx, and provides dual-layer auditing for estimation and rendering along with PDF preview. Installation: dsh plugin --profile default add github:John-20260101/zgs-ppt; it can also be installed via SkillHub @user_3b69e11c/zgs-ppt.
Read Moredsh-task-checkpoint-guard: Local Checkpoints and Bounded Recovery for DSH Long Tasks
Introducing the community plugin dsh-task-checkpoint-guard (v0.1.0, MIT): It writes local JSON/Markdown checkpoints for long-running tasks under the DSH 0.2.1-rc.3 environment, with sanitization and no chain-of-thought saved. It classifies network/timeout/service failures, defaults to checkpoint-only, and optionally performs bounded recovery in probe or deferred mode with re-arm Goal. Includes npm pack installation, cordis.patch configuration examples, and checkpoint directory verification commands, along with explanations of the fail-closed policy, goal-round-driver dependency, and process permission considerations.
Read Moredsh-easyrestart: Adds a native "Restart" menu to DSH Desktop
dsh-easyrestart is a DSH client plugin maintained by jkrhug. Under the desktop profile only, it adds a native Restart item above Quit to the tray right‑click context menu and the Application menu of the Windows DSH Desktop, solving the issue where closing the window merely hides to the tray and requires quitting and manually reopening to take effect. Install with dsh plugin --profile desktop add dsh-easyrestart; after installation you must fully quit and then reopen Desktop. You can manage the patch via /easyrestart, dsh_easyrestart, and the plugin management page. Implementation-wise, when permissions allow it modifies lib/main.js inside the installation directory’s resources/app.asar; it does not connect to the network and does not report telemetry. Desktop upgrades will overwrite the asar, requiring the patch to be reinstalled. MIT licensed; current package version is 0.2.0.
Read Moredsh-update-guard: DeepSeek Harness Pre-Update Compatibility Check and Startup Self-Healing
dsh-update-guard is a DSH desktop workflow plugin published by jiuyuan-yi: when an update ready for installation is detected, it unpacks and compares the interface surface with the version manifest, and for missing client UI members it can optionally automatically lock via the configurable cordis.patch.yml; the host inject service only prompts. An optional wscript launch wrapper handles backup and restoration in cases of launch failure or when plugin configuration is cleared. Requires Node 18+ and Windows, and operates offline. This article, based on the GitHub README, introduces installation, the CLI, and configuration options.
Read MoreSiemens WINCC Assistant: Error Decoding and Configuration Assistance Skill for WinCC Engineers
The Siemens WINCC Assistant targets automation engineers, electrical engineers, and programmers using SIMATIC WinCC Classic V6/V7 and TIA Portal. It provides deterministic assistance for five high-frequency pain points: decoding errors with 18 built-in official HResult codes and 6 troubleshooting checklists; converting Excel/CSV variable tables into importable TIA Portal Hmi Tags XLSX files (solidifying official 15-column headers and freezing the first row); generating Alarm Logging tables; providing C/VBScript action templates (including type conversion and COM release syntax); and offering four checklists (runtime startup, commissioning, backup/archive, system health) along with 10 lessons learned. Built entirely with the standard library and requiring zero pip dependencies, it runs on Python 3 on Windows/Linux without an API key. Released and maintained by community developer user_51a923b7, Version 1.0.9, categorized under it-ops-security, with a SkillHub score of 100,000 and 1,319 downloads.
Read MoreFun Math Connect: Integrating Core Math Mini-Program Features into AI Dialogue
Fun Math Master is a community-sourced Agent Skill (@user_f7c670f8/qwsxt, category: education, version 1.0.1, requires_api_key=false) that integrates the core functionalities of the "Fun Math Master" WeChat mini-program into AI conversations. It covers modules including mental arithmetic practice, puzzle games (e.g., 24 Game, Sudoku), mathematical exploration (e.g., Chicken and Rabbit in the Same Pen, General's Drinking Horse), fun riddles (graphical and logical reasoning), an AI learning companion, error management, study planning, unit conversion, and user growth (membership system, thinking levels, learning reports). It targets students, parents, and teachers who wish to learn mathematics through dialogue. This article describes its positioning, functional scope, installation and activation methods, typical usage, and applicable limitations, while honestly flagging installation commands that could not be verified.
Read Moredsh-plugin-background-compaction: Reducing DSH context compaction turn freeze from 16 seconds to 7 milliseconds
This article introduces dsh-plugin-background-compaction (v2.0.1), a DSH plugin maintained by Jisiyong-A. It replaces compaction-basic by pre-computing LLM summaries in the background when the threshold reaches the 72% watermark. Upon triggering, it commits directly, reducing compression blocking from 16 seconds to 7 milliseconds in benchmarks, with approximately a 2.2% increase in billable input tokens. The article covers the architecture, installation commands, cordis.patch configuration, differences from pruner/dsh-dcp, fail-open behavior, and usage considerations.
Read MoreHigh School Error Photo Collection Diagnosis: From Photo-taking to Error Collection and Weak Point Diagnosis
This article introduces the community Skill “High School Mistake Photo Collection & Diagnosis”: users can upload photos of exam papers or exercise sheets, or manually enter wrong questions; the system aggregates and categorizes them by subject and knowledge points, generates printable mistake collections or Anki review cards (including sanitization with injection prevention), and outputs a diagnosis report on weak areas with recommendations of similar questions. Version 1.2 adds an image recognition pipeline and an exam-point knowledge base (14 exam points across 6 subjects); version 1.1 adds a standalone mistake collection tool. The article describes the core features, how to install and enable the Skill, trigger words and typical usage, and provides applicable scenarios and limitations.
Read Moredsh-model-rank: Multi-list model comparison and local usage in the DSH sidebar
dsh-model-rank is a DSH client plugin (v0.2.1, MIT) maintained by Jingxuan-WH that provides a "Model Rankings" section in the sidebar. It displays six separate leaderboards: Arena blind tests, Artificial Analysis Intelligence Index, LiveBench 2.0, and OpenRouter usage and market share, while reading `cordis.patch.yml` to highlight integrated models. The local integration page features a token dashboard and cost estimation, aggregating data from session logs in read-only mode without uploading any data. The panel loads instantly from cache, while the host fetches public rankings according to preset rules; the composite intelligence and capability rankings require an initial online refresh on the local machine. It supports sorting, detail drawers, and CSV export without requiring an API key. Compatible with DSH 0.2.x and Node.js ≥ 22.15, it can be installed via the settings page or by running `dsh plugin add`.
Read Moredsh-pet-tts: Reading Desktop Pet Speech Bubbles with Edge Neural Voice
dsh-pet-tts is an online tool plugin for DeepSeek Harness maintained by jimmyfenwick. It only reads the .dsh-pet-bubble elements of dsh-pet in the Web UI, and the host synthesizes mp3 files using edge-tts with Microsoft Edge's online neural voices. It supports preset voices, a console API, and an /dsh-pet-tts HTTP route. On first use, a Python virtual environment is created in ~/.dsh/dsh-pet-tts. It does not read the assistant's main text. Requires Python 3.9+, Node 20+, and an internet connection. Installation: dsh plugin --profile <profile_name> add github:jimmyfenwick/dsh-pet-tts.
Read More