Using Official OAuth to Log into ChatGPT (Codex) in DSH
dsh-codex-login is a DSH plugin maintained by HarryLi-7. It drives the existing ChatGPT OAuth flow on pi-ai and dsh-authorization through the Agent tool codex_login, writes the credentials to llm-pi-ai/openai-codex, and enables the GPT of the openai-codex provider to serve as the Harness primary model. The article explains the background of the plugin's existence, tool actions, credential fields, the model catalog configuration syntax for an empty provider, as well as the link installation steps and environment requirements (DSH 0.2.1-rc.5) under the desktop profile. It also reminds users about the callback port 127.0.0.1:1455, using a browser on the same machine, and reviewing the source code before installation.
Read MoreTalent Review 9-Box Grid → Action Converter: Translating Assessment Results into Action
Based on the Performance x Potential nine-box grid model, this Skill transforms talent review results into a differentiated action plan. It maps the nine-box grid to a four-dimensional action matrix covering development, motivation, and retention (including "what not to do"). It features a four-step facilitation script for calibration meetings, early warning signals for high-potential turnover, a quarterly closed-loop tracking system with four checkpoints, six exception handling scenarios, and a complete diagnostic case study of a 7-person technical team in a 40-person startup. Community release, Version 1.0.1, Category: business-ops, No API Key required.
Read MoreProject Management & Scheduling Skill: Project Planning, Scheduling, and Progress Control
"Gut-feeling scheduling and verbal progress updates are among the most common issues in team project collaboration: plans are scattered across documents and spreadsheets, milestones lack oversight, and risks are not identified in advance. The Skill introduced below consolidates project planning, scheduling, risk, and progress management into a single workflow, supporting both Agile and Waterfall methodologies. ## What is this "Project Management & Scheduling" is an Agent Skill released by "Kunlun Growth" (coordinates `@kunlungrowth/biz-project-management`, current version 1.0.0). It is designed for project planning, scheduling, risk, and progress management. It supports both Agile and Waterfall modes and includes milestone control. Classified under business-ops, this Skill..."
Read Moredsh-fitness: Tracking Weightlifting Workouts with Local SQLite in DSH
dsh-fitness is a community plugin for DeepSeek Harness maintained by hardlic. It registers a fitness_log tool with the Agent (supporting eight actions in total: logging workouts, viewing history, generating weekly reports, and creating training plans). An optional UI dashboard can be installed to display a weekly calendar, activity logs, and suggestions in the left sidebar. Data is local-first, stored in $DSH_HOME/dsh-fitness/fitness.sqlite, with no network access or account required. Installation requires an existing DSH and plugin_manager setup; install via clone-then-install_bundle or by adding the local directory through the plugin page. The dashboard requires separately installing the ui directory. MIT licensed, Node >=22.5, current version 0.3.1.
Read MorePython: Agent Skill with Configured Activation Prerequisites for Python Projects
Python is an Agent Skill designed for Python projects. It is located at `@org-eyw2ohcx/sammcj-agentic-coding-python-development`, version 1.0.0, categorized under dev-programming, published by RealQ, and marked as enterprise source. Its core is a trigger-based introduction: "You MUST activate this skill when working on Python projects", which means that when a Python project is identified, this Skill should be activated first to allow the Agent to work within a Python context. The label `requires_api_key=false` indicates no additional key is needed, and it is based on the general SKILL.md format, making it compatible with Agent environments that support this format. The fetched data does not provide installation commands, configuration fields, or reproducible examples, so these details are not included here. The SkillHub page shows a rating of 100000, 985 downloads, and 0 installations, indicating very low actual installation volume. Its effectiveness needs to be verified by users themselves. This Skill is suitable for users handling Python projects in environments that support SKILL.md and want the Agent to explicitly identify project types first.
Read MorePlaywright MCP Browser Automation: Delegating Browser Operations to Agent Workflows
This is a browser control Skill based on official Playwright, designed for agent workflows. It encapsulates browser operations into MCP tools callable by Large Language Models, allowing agents to perform actions such as opening pages, clicking, inputting, and reading page content via tool calls, without the need to write custom scripts for every task. Verified Facts: Handle @user_3c6cb52e/playwright-mcp-browser-automation, Version 1.0.0, Source: community, Category: ai-agent, Tags: requires_api_key=false. As the official page retrieval failed, the installation command, tool list, configuration fields, and typical usage have not been verified; this is explicitly stated without fabrication.
Read MoreExpert in Hand-drawn Colored Pencil Childlike Illustration Posters: Converting Children's Poster Requirements into Structured Prompts
This is a SkillHub community skill located at @user_442551c3/colored-pencil-kids-illustration-poster-expert-600029, version 1.0.0, categorized under design-media. It caters to poster requirements for children's picture books, parent-child activities, early education, fairy tales, Children's Day, and birthday parties. It organizes descriptions based on multi-layered colored pencil stroke textures, soft and cute anthropomorphic characters, low-saturation warm tones, picture book narrative scenes, and life details to output complete prompts suitable for AI image generation tools. While the metadata indicates requires_api_key=false, no verifiable installation commands or configuration fields are provided. It is suitable for scenarios such as picture book covers, event posters, holiday cards, and children's brand materials. Please note that its style is restricted to hand-drawn colored pencil childlike aesthetics, and it does not generate images directly.
Read MoreDSH-Paper-DeepRead: Running the Literature Deep Reading Repository and Skills in the Harness Sidebar
DSH-Paper-DeepRead is a dsh Web plugin (v1.5.2) maintained by HaoKuo. A single installation embeds a static site for a curated literature library and the paper-deepread-html skill, offering sidebar browsing, managed PDF uploads, model tools such as paper_deepread, and report management. It enables Agents to execute ingest/build operations via uploads/reports tools and supports an optional siteUrl for automatic deep reading integration with standalone sites. Requirements include dsh 0.2.0-rc.2–0.2.1-rc.3, Node ≥24, and a web profile; installation involves downloading the tgz from Release and running `dsh plugin --profile web add`.
Read MoreTesting Expert: QA Skills Covering Requirement Review to Production Launch
Test Expert (@indiv-skill/test-expert, 1.0.0, category ai-agent, source community, publisher user_e9185b15, requires_api_key=false) is a software testing and quality assurance Agent Skill. Trigger scenarios include test strategy and planning, test case design (equivalence partitioning, boundary value, decision table, scenario, cause-effect graph, Pairwise, etc.), functional, API, performance, security, compatibility, and mobile testing; automation (pytest/Playwright/Cypress/JUnit/Postman); quality metrics and defect management; test reporting; release gates; shifting testing left (TDD/BDD/contracts) and right (gray release/chaos/observability); AI/LLM application testing; and test data and environment management. Covering the entire pipeline from requirements review to production deployment, it can directly generate test cases, analyze coverage, and triage defects, producing reusable templates for test plans, cases, defects, and reports.
Read Moredsh-small-toolkit: macOS Wake Screen and Task Completion Alerts for DSH Runtime
dsh-small-toolkit is a DeepSeek Harness client plugin (MIT license, v0.3.0) maintained by HanYu-me, effective only on macOS. Its keepAwake feature uses caffeinate during dsh runtime to prevent screen dimming and system sleep, while taskNotify sends a system notification when a task round in the main conversation completes—clicking it switches back to DSH. It supports hot-toggle settings in settings.json and fine-grained configuration via cordis.patch.yml. To install, run `dsh plugin --profile web/desktop add github:HanYu-me/dsh-small-toolkit`, then restart dsh after installation.
Read MoreIn-depth Industry Penetration Research: Conducting a Penetrating Analysis of an Industry through Four-layer Progression
"Industry Deep Dive" is a community Skill on SkillHub (@user_b7d91cbb/industry-deep-dive, v1.0.0, Professional category) that adopts a universal SKILL.md format and requires no API Key. It transforms vague requests such as "research a certain industry" into a four-layer progressive analysis framework: Macro Panoramic Positioning → Mid-level Chain Decomposition → Micro Competitive Assessment → Future Trend Deduction. Supporting both Chinese and English trigger words, it delivers complete penetrative analysis ranging from cognitive foundation building to decision-making reference. The article covers its positioning, four-layer logic, installation and activation methods, typical usage (Chinese and English examples), applicable scenarios, and precautions. It is suitable for scenarios such as pre-investment due diligence and cognitive foundation building before entering new markets.
Read MorePrompt Optimization (Integrated SuperPowers Edition): Adjusting Constraint Density with Six Keyword Prefixes
Introducing the community Skill "Prompt Optimization (Integrated SuperPowers Edition)" (Constraint Density Adjuster), located at @user_07f065a9/constraint-density-adjuster, version 1.0.6, categorized under ai-agent, sourced from the community. The core is a six-keyword command system: predictable-run/predictable-root/predictable, which respectively tighten constraints and execute directly, find root causes, and intelligently mix tightening; unpredictable-run/unpredictable-root/unpredictable, which respectively loosen constraints and explore, challenge assumptions, and intelligently mix loosening. To enable it, execute a command in a workbuddy conversation, pasting the Pre-processing Rules before Core Truths in SOUL.md (appending to the end if the file is empty). When a message prefix matches, processing stops following the original text and instead invokes Skill(constraint-density-adjuster) to transform the prompt before execution. The official page shows a rating of 100000, 940 downloads, 0 installations, and no API Key is required. The article provides installation configuration blocks, typical prefix usage examples, applicable scenarios, and precautions.
Read Moredsh-update-center: Check versions, view release notes, and one-click upgrade in DSH Web
dsh-update-center is a DSH client plugin maintained by HansyLin-xCoFe, providing an update center within the Web interface. Under Settings > General, it displays the current and latest versions, allows checking for updates, and offers one-click upgrades; a popup appears in the bottom left corner when a new version is available. Update highlights are derived from the Chinese paragraphs of the dsh-v* Release in deepseek-harness. Opening the page automatically checks the npm latest version (Host caching for 10 minutes). The one-click update process uses a detached assistant to write the schedule, exit the Host, perform a global npm installation, and then restart the dsh web. The installation spec is github:HansyLin-xCoFe/dsh-update-center, requiring DSH >= 0.2.1-rc.2 and Node 18+. MIT License.
Read MoreE-commerce Copywriting (Taobao/JD/Pinduoduo): Introduction to the Multi-platform Product Copywriting Skill
This section introduces the community Skill for e-commerce copywriting (Taobao/JD/Pinduoduo), with the coordinate @user_4f99260d/ecommerce-copywriter-v2. Version 1.0.0, from the content-creation category on SkillHub, published by user_4f99260d, it requires no API Key. This skill generates product titles, main image copy, detail page copy, selling point extraction, promotional activity copy, and live streaming scripts for Taobao, JD, Pinduoduo, and Douyin E-commerce based on product information, adapting to platform styles and word count limits. The main text covers what it is, core features, installation and activation, typical usage, applicable scenarios, and precautions. Due to the failure to scrape both the official page and the API during this session, the installation commands and tool differences have not been verified; only the official address and coordinates are provided for cross-checking, and no configurations or cases have been fabricated.
Read Moredsh-quote-annotate: Select paragraphs in conversations, annotate inline, and send with the next message
dsh-quote-annotate is a DeepSeek Harness client plugin maintained by HansHuang2683, compatible with the Web/Desktop Web UI of DSH 0.2.1-rc.3+. Users can select any text within the transcript to annotate; after saving, it highlights with numbered markers and summarizes them above the composer. When sending a message, the original text, source, and annotations are collapsed into a 【我的批注】 block and passed to the model along with the message, alleviating the problem of unclear references in long conversations. Installation is performed via plugin_manager's install_bundle pointing to the GitHub repository, and it displays as "Selection Annotations" in the Plugin Manager.
Read MorePractical AI Assistant for Communications Engineering: Put Field-Ready Experience to Work with Your Agent
The "Communication Engineering Practical AI Assistant" is based on the "Weak Current Intelligent AI" knowledge base, targeting frontline communication engineering scenarios. It packages optical attenuation parameters, compensation unit prices, cutover procedures, Excel/Word templates, and pitfall avoidance notes from transmission, wireless, line, core network, and settlement into Agent-callable content. It covers five practical modules including Huawei/ZTE equipment troubleshooting, base station surveying, line relocation compensation, and construction pre-settlement, with requires_api_key=false. This article explains its positioning, core functions, activation methods, typical usage, and applicable boundaries, and honestly marks installation commands and configuration fields that could not be verified from official materials.
Read MoreCallAI SEO MCP: Enabling Real-time Search and SEO Data for the DSH Agent
This article introduces the @callaiapi/seo (callai-seo-mcp) DeepSeek Harness plugin maintained by hanshengcc. Utilizing MCP, it provides 11 tools for Google/Bing SERP analysis (including AI Overview citations), keyword difficulty and long-tail variations, competitor gap analysis, as well as backlink and page auditing. Data is sourced from DataForSEO and CallAI, with pay-as-you-go billing. It covers the official dsh plugin installation command, API Key and cordis.patch.yml configuration, self-check methods, and typical query examples from the README.
Read MoreJiandaoyun Auto Assembly Assistant: Batch Form Skeleton Construction and Common Error Avoidance
Introducing the JianDaoYun Auto-Setup Assistant: Released and maintained by @user_cacd69de (Source: community, Category: office-efficiency, Tag: requires_api_key=false). It handles two main tasks: First, it explains the platform methodology of JianDaoYun regarding forms, workflows, dashboards, and Open APIs for solution consulting and software selection. Second, utilizing Playwright and CDP, it takes control of the logged-in backend to batch create apps and ERP form skeletons. It automatically adds master table fields, sub-forms, and sub-fields with precise naming. Furthermore, it integrates practical pitfall avoidance measures into the workflow, such as the "jump-off" creation method, making sub-field boxes read-only, avoiding accidental elimination of the left panel, preventing accidental modification of sub-fields rolling out of view, preventing continuous table creation pollution, utilizing precise blacklists for real tables, enabling breakpoint resumption and same-name self-healing, and implementing watchdog self-waking. When cleaning up test forms, it relies on precise blacklists to avoid accidentally deleting real user tables; property configuration is left for the user to complete. The article also outlines applicable scenarios and risk points, noting that discrepancies exist regarding ratings and download data in the SkillHub snapshot, and real-time page data should be taken as the standard.
Read MoreBeihang University - Research PPT Skill: Converting Papers, Notes, and Experiment Records into Standardized XeLaTeX Beamer Presentations
Introducing the community Skill @user_9b3dfcf6/buaa-beamer (version 1.0.2, category: office-efficiency): a XeLaTeX Beamer workflow tailored for research and study scenarios at Beihang. It first understands papers, notes, experiment logs, or existing TeX sources, then uses a minimal interview to confirm the scenario, duration, audience, and key points, and afterward generates, compiles, visually inspects, or reviews slides for settings such as group meetings, paper seminars, close reading of literature, research progress updates, courses, or technical sharing. It provides compilation, TikZ, review, and quality-control capabilities, supports Chinese via XeLaTeX, adopts a default Beihang blue-and-white academic style, and prioritizes an official template when the user provides one. The page is marked as requires_api_key=false, so no API key is required.
Read MoreAutomotive Full Lifecycle Expert: Linking vehicle selection, usage, maintenance, and content into a callable chain.
Introducing the community Skill "Automotive Lifecycle Expert" (`@user_cae83fba/car-lifecycle-media-expert`, v3.1.0, source: community, category: content-creation, no API Key required). It integrates vehicle selection (profile scoring, anti-recommendation, TCO, and three thresholds), usage costs and energy type selection (BEV/PHEV/EREV/ICE), fault maintenance (three-tier quotes and dealership scripts), content creation across 10 platforms (KOL styles and reviewer tone imitation), field-level source confidence with eight-dimension validation (including consistency), pre-publication mandatory gates with return attribution, and task state machines with interruption recovery into a unified `SKILL.md`. Optional integrations include libraries, RAG, REST, and MCP. The article outlines core capabilities, installation and activation methods (no command-line installation available on the page; requires manual placement into the client's Skills directory, with variations across three client types requiring empirical testing), invocation examples, and applicable boundaries (not intended for real-time inventory transactions, substituting professional maintenance, financial loan decisions, or autonomous driving regulatory rulings). At the time of scraping, the page showed a rating of 100,000, 1,055 downloads, and 0 installations; independent verification is required before adoption.
Read MoreHong Kong Local Recommender: Adding a Local Friend's Perspective as a Knowledge Source for Agents
The Hong Kong Local Recommender is a Skill designed for queries regarding local lifestyle and tourism consumption in Hong Kong. Located at `@indiv-tools/hongkong-local-recommendations`, Version 1.0.3 (Community sourced, no API Key required), it provides dining, accommodation, and entertainment recommendations in the voice of a genuine local resident. Each suggestion is tagged as "Frequented by Locals," "Tourist-Friendly," or "Mixed," ensuring authentic advice rather than generic guidebook content. The tool addresses six categories of inquiries: dining recommendations, attractions, convenient accommodation areas, MTR navigation, tourist traps to avoid, and 3-day itinerary planning. Outputs include overviews, recommendation lists, route suggestions, etiquette tips, and pre-departure verification reminders. Place and shop names are preserved in Traditional Chinese with Cantonese Romanization, while prices and business hours are strictly marked as "subject to the latest updates." Trigger words include Hong Kong recommendations, Hong Kong food, Hong Kong guides, Hong Kong itineraries, avoiding Hong Kong scams, and Hong Kong Cha Chaan Teng. This Skill is exclusively for questions regarding local living and tourism consumption in Hong Kong; it is not intended for other cities, translation or rewriting tasks, real estate investment, school selection, or financial report analysis.
Read Moredsh-breezetts: Using Breeze TTS to Read Aloud Assistant Replies in DSH
The DSH community plugin dsh-breezetts, maintained by hangdingding, integrates Breeze TTS within DeepSeek Harness sessions to read aloud the assistant's text replies. Installation is available via the Git source github:hangdingding/dsh-breezetts or the Release package local-breeze-voice-x.y.z.tgz. You need to prepare an API Key starting with brz_ and a voice_id starting with voc_ from Breeze Studio, both configured in the plugin panel without modifying the DSH configuration file. It is suitable for developers who need to listen to assistant replies; please review the source code before installation and note that the plugin runs with DSH process permissions.
Read MoreCommuting House-hunting Assistant: Transforming Commute Constraints into Actionable Search Reports
# Introduction If you are developing a real estate search Agent, users most often mention requirements like "easy commute," "within 45 minutes by subway," or "near the office." Without translating these requirements into concrete criteria, the Agent might easily switch back and forth through property listings and ultimately deliver a bunch of links without being able to clarify which property is actually feasible to live in. The following is an introduction to the "Commute-Friendly Home Search Assistant," a community Skill on SkillHub. Targeting homebuyers and real estate agents, it transforms "commute" from a vague requirement into verifiable data—including routes, times, and scoring criteria—for each property. # What is This The "Commute-Friendly Home Search Assistant" is an Agent Skill based on the universal SKILL.md format. It is published by user_1266dc16, sourced from the community, and is at Version 1.
Read MoreMetallurgical Engineering Quality Planning Assistant: Ensuring Executability
This Skill, published by "Peng Ge Liao Gongcheng AI" (user_833fca90) in the SkillHub community, is designed for new, renovated, and expanded metallurgical engineering projects. It assists project managers, chief engineers, and quality directors in generating, completing, reviewing, and upgrading the *Metallurgical Engineering Quality Planning Scheme*. It features a built-in metallurgical process unit system (covering raw materials, sintering, coking, ironmaking, steelmaking, hot rolling, cold rolling, forging presses, and energy efficiency & green buildings), a quality risk checklist, and planning frameworks for individual commissioning, load-free linkage commissioning, and hot load commissioning. The tool outputs a complete 17-section scheme structure and supports the creation of specialized tables such as goal decomposition matrices, risk analysis charts, and commissioning planning schedules. No API Key is required for usage. Users only need to provide the project type, construction scale, main process units, quality objectives, and current status to generate a draft. This Skill does not replace the approval of construction organization designs, special acceptance conclusions, or third-party inspection reports; manual review is required for significant matters.
Read MoreBreezeTTS for DSH: Read Assistant Responses Aloud
breezetts-dsh is a Breeze speech plugin developed by hangdingding for DeepSeek Harness. It prioritizes device login via the Breeze CLI, falling back to the Breeze API when unavailable. It supports text-to-speech buttons, automatic reading, voice search and design, speech rate and pronunciation guidance, and registers `breeze_speak` to allow models to synthesize customized audio. The installation identifier is `github:hangdingding/breezetts-dsh`; in browserless environments, `breeze login --no-browser` or `BREEZE_API_KEY` can be utilized.
Read More