Odyssey | Life Designer: Conducting a Life Simulation with Stanford's "Odyssey Plan"
"Odyssey | Life Planner" is a community Skill based on Stanford's "Designing Your Life" Odyssey Plan methodology, located at @indiv-witt/odyssey-plan, version 1.0.1, categorized as life-service, published by user_d241203e, and requires no API Key. It drives a full life simulation through a five-stage in-depth dialogue: first diagnosing the current situation, then uncovering inner conflicts, then exploring three parallel life paths, stress-testing the costs of each path, and finally designing minimum-cost experiments, outputting a life simulation report. The Skill is actively triggered and covers scenarios such as feeling lost about life direction, career confusion, switching careers vs. staying put, major life choices, not knowing what to do next, life design, and career crossroads. Even a request like "help me think about what to do next" will trigger it. No independent installation command is provided in the official docs; as a Skill based on the generic SKILL.md format, it can be placed according to the skills directory conventions of Agent clients. It is intended for people who are directionally lost, hesitating over major choices, or seeking to systematically sort out their lives. Note that this is a conversational guidance tool; output quality depends on the user providing truthful and specific information, and it does not replace professional career counseling or psychological diagnosis. As of retrieval: 1,198 downloads, 0 installs, making it a newly released community Skill.
Read Moredsh-cat-diary: Four-Column Diary and Affection Levels Enable New DSH Sessions to Remember Yesterday
dsh-cat-diary is a memory plugin developed by AsterLuo for DeepSeek Harness. It captures all conversations on a natural daily basis; the model writes the first three sections, while the code generates the fourth. Upon startup, it idempotently backfills any outstanding diary entries, and automatically injects a "yesterday's quick-read card" into new sessions. The affinity sub-plugin evaluates status during the agent/pre-step phase of each turn, ensuring unique ledger writes, with optional read-only display via a sidebar. Based on the official README, this article outlines installation commands, Cordis configuration, CLI usage, and storage paths, while detailing usage and security precautions for environments without a web panel.
Read Moredsh-channels: Discord-style Channel Collaboration for DeepSeek Harness
dsh-channels is a DSH web client plugin maintained by Asheep233. It provides a channel hub and React panel within the dsh web process: public subchannels, append-only immutable history, @mention wake-up and steer/resume sessions, and exposes a loopback HTTP API, 21 channel tools, CLI, and MCP. It is designed for local multi-harness and same-channel collaboration. Installation requires DSH 0.2.1-rc.x and Node.js 20+, using the command: dsh plugin --profile web add github:Asheep233/dsh-channels.
Read MoreFinancial Analysis Report Writing Smart Assistant: Writing Framework for the "Financial Condition Analysis" Chapter in Fixed Income Reports
This is the community Skill "Investment Banking Wit - Financial Analysis Report Writing Assistant" on SkillHub, located at @user_4398eccb/fixed-income-financial-analysis. Published by user_4398eccb, it is Version 1.0.7, categorized as "professional," sourced from the "community," and tagged with requires_api_key=false. Targeting fundraising prospectuses and due diligence reports for fixed-income instruments such as corporate bonds, enterprise bonds, ABS, REITs, PPNs, and MTNs, it provides a general writing framework for the "Financial Status Analysis" chapter. The official description lists four core capabilities: a 13-paragraph chapter skeleton, definitions for 14 financial metric formulas, a debt repayment analysis framework, and writing standards. The prompt directions focus on writing financial analysis chapters, metric formula definitions, debt repayment analysis, and analysis of the issuer's financial status. The public page does not provide installation commands, configuration fields, or reproducible calling examples. This article leaves these details blank accordingly and does not fabricate commands or samples; specifics should refer to the actual Skill files.
Read Moredsh-balance-ru: Harness sidebar balance and daily consumption, displaying rubles at Central Bank of Russia exchange rates
dsh-balance-ru is an unofficial Russian fork of @pipilu123566/dsh-balance (MIT, 1.2.0), designed for DeepSeek Harness Desktop 0.2.1-rc.5. It aggregates multi-provider balances and daily token consumption estimates in the sidebar, displaying approximate ruble values for USD/CNY based on the Russian Central Bank's daily exchange rates; a Russian dictionary is built-in. To install, download the release tarball (tgz); remove the upstream plugin first if necessary, then run `dsh plugin --profile desktop add ... --ignore-scripts`. Classified as a network tool, it requires internet access to fetch exchange rates (cached for 6 hours).
Read Moreself-improvement-agent-pro: Turning AI's Corrections into Cross-Session Experience
self-improvement-agent-pro is an AI self-improvement Skill released by the community (v1.1.1). Classified as an ai-agent, it requires no API Key. It writes every correction, error, and missing requirement into the project's learnings directory; high-frequency experiences are upgraded to CLAUDE.md/AGENTS.md, while stable experiences can be extracted into new skills. It provides capabilities for error correction logging, experience accumulation, cross-session memory, skill extraction, log retrieval, and weekly review. It can be triggered by both technical commands and natural language. This article explains core features, enablement methods, typical usage for recording, retrieving, and reviewing, as well as the boundaries and considerations of project memory. Installation commands should refer to the official page.
Read Moredsh-plugmart: DSH Left Sidebar Plugin Marketplace
dsh-plugmart is a DSH client plugin maintained by archiecheung. It provides a Plugin Market entry in the left sidebar of the Web interface. Features include 250 ms debounced search, Sorting by "Users" based on star popularity, filtering via 16 popular tags, pagination and a details sidebar, and one-click copy of install commands. It bundles snapshots of ~700 market.json entries for offline browsing and, when online, can refresh the community index at 2bingling.github.io and cache it to DSH_HOME. To install, run npx @deepseek-ai/dsh plugin --profile <profile> add dsh-plugmart; after installation, restart DSH. It is an MIT-licensed, open-source project that does not collect user data.
Read MoreClinical Knowledge In-depth Decoding Expert: Knowledge Extraction from Medical Texts and Clinical Protocol Analysis
Introducing a community Skill on SkillHub: Clinical Knowledge Deep Decoder (User ID: @user_29051052/clinical-decoder, version 1.0.0; category: Professional). This Skill scopes tasks to knowledge extraction and analysis in medicine, health, and biology, and provides clinical guideline interpretation. It is labeled requires_api_key=false and does not require an additional API key. The article explains its positioning, two verifiable core functions, prerequisites for enabling it via the universal SKILL.md format (as the source material does not provide official installation commands and configuration fields, none are fabricated here), typical usage patterns, and review considerations. It also notes that the Skill comes from a community source and does not replace clinical judgment.
Read Moredsh-prompt-templates: Manage prompt templates and thinking shortcuts in DSH
Apzx7 maintains the DSH client plugin `dsh-prompt-templates`, offering 34 built-in Chinese-English bilingual templates, collapsible categories, search and filtering, favorites, and optional AI-powered categorization. Data is stored in `~/.dsh/prompt-templates.json`. It features 9 AI Coding mindset shortcuts and a Liquid Glass material interface. Installation: `dsh plugin --profile web add dsh-prompt-templates`. Compatible with DSH 0.2.x and Node.js ≥20. Smart categorization reuses the DSH LLM to send template summaries to configured models; this feature can be disabled to maintain offline usage.
Read MoreOn Calling Image and Video Generation Interfaces Such as Nano Banana, Gemini, Jimeng and Suno within Skills
MXAPI is a community-sourced Skill (Coordinates: @user_22f85b9d/mxapi, Version 1.0.4, Category: design-media) that integrates with the MXAPI Open Platform (open.mxapi.org) to consolidate multiple active interfaces for image, video, and music generation into a single entry point. It covers Nano Banana, Gemini image generation, GPT-Image, OpenAI-compatible image generation via GPT 2.5 and Nano/GPT, as well as video generation via Jimeng, Seedream, Seedance, and MiniMax H3. Additionally, it supports Suno for songs, BGM, and instrumental music, along with image-to-image, image editing, and first-last frame video generation. It also allows checking credit balances and API pricing. An API Key is required. The article introduces its positioning, core features, installation entry points, typical usage triggers, and precautions.
Read MorePaper Figure Code Generator (Ready-to-Use Publication-Grade Scripts): Translating publication-grade figures into executable scripts
Introducing the community Skill “Paper Figure Code Generator (Ready-to-Use · Publication-Grade Figure Scripts)” (@user_af163eaa/z-pub-figure, version 1.0.1, category dev-programming, tag requires_api_key=false). Targeted at paper illustration scenarios, it generates matplotlib / ggplot2 scripts with a default orientation toward high DPI, vector-style aesthetics, and font compliance. Users run these scripts locally to export PDF/SVG outputs, which differ from the output form of interactive HTML visualizations. The article clarifies the capability boundaries (only scripts are delivered, no image hosting is provided), the verifiable parts of the installation information (the materials do not provide installation commands, and the official page is currently inaccessible, so no steps are fabricated), generic call flows and script form examples marked as illustrative, as well as applicable scenarios and limitations (users must prepare their own runtime environment and fonts, vector style does not equal automatic aesthetics, interactive scenarios are not applicable, and the functionality of version 1.0.1 may be relatively narrow).
Read Moredsh-bili-statistics: Comparing planned and actual Bilibili viewing
dsh-bili-statistics is a DSH web plugin maintained by apiaoapiao that fetches personal watch history via Bilibili's official API, performs local statistical calculations with optional model-generated overviews and plan completion reports. It supports Markdown/HTML reports, scheduled tasks, and Bilibili private message and email notifications. The installation command is `dsh plugin --profile web add dsh-bili-statistics`, requiring Node.js 22+ and QR code login. Based on the GitHub README and package.json, this article introduces features, settings pages, the bili_intent_* toolchain usage, and usage notes.
Read Moredsh-lazy-toolbox: A Lazy Toolbox for On-demand Hot-loading of Skills and MCPs
By default, DSH sends all skills and the complete tool definitions of every MCP into each request. The more you load, the more tokens and tool-selection noise increase. dsh-lazy-toolbox moves skills/MCPs out of global definitions and keeps only a tiered catalog in the system prompt, with a one-line description of each item’s purpose. Within a session, a persistent toolbox lets you load/unload tool definitions on the fly, supports on-demand fetching of skill outlines (fetch specific sections as needed), plugin-based tool-group hosting, and toggle switches on the settings page. Based on the official README, this article introduces installation (npm, GUI, wire-profile), typical toolbox operations, and usage considerations.
Read MoreTax Bureau Login Assistant: A Configuration-Driven Skill for Assisting Login to the Electronic Tax Bureau
# Preface In their daily operations on the Electronic Tax Bureau, financial personnel often encounter three types of issues: 1. **Repetitive Data Entry:** They must re-enter the web address, full company name, abbreviation, Unified Social Credit Code, mobile number, and password for every login, making manual entry prone to errors. 2. **Multi-Enterprise Management:** When responsible for multiple companies simultaneously, they need to switch back and forth between different credentials. 3. **Security Verification:** Security verification steps (such as image CAPTCHAs and SMS verification codes) cannot be handled by automated scripts and must be completed manually. The "Tax Bureau Login Assistant" introduced below provides a configurable and switchable auxiliary skill designed to address these repetitive operations. ## What is this The "Tax Bureau Login Assistant" is a configuration-driven, multi-enterprise auxiliary skill designed for financial personnel to assist with logging into the Electronic Tax Bureau. It is identified by the coordinate `@user_b5a96cf7/etax-login-assistant` and is developed by `user_b5a96c...`
Read MoreJingfang-medicine: Decomposing Classical Formula Pattern Differentiation into Domain Workloads across Six Domains and 28 Tasks
jingfang-medicine is a "Professional" category Skill from the "Community" source on SkillHub. Version 1.0.0, located at @user_4c004f18/jingfang-medicine, requires_api_key=false. It is positioned as a workload application in the field of Jingfang (Classical Prescription) medical syndrome differentiation and prescription simulation, breaking down a single Jingfang diagnosis and treatment session into 28 tasks across 6 domains: information collection for the Four Diagnostics, Six-Meridian syndrome differentiation analysis, selection and modification of classical prescriptions, dosage and decoction administration medical orders, autonomous medical case study, and therapeutic evaluation iteration. It features built-in Jingfang safety red lines, a knowledge base for prescription-syndrome correspondence, and models of renowned physicians' syndrome differentiation reasoning. Its capabilities derive from the original texts of the *Treatise on Cold Damage Disorders*, Hu Xishu's Six-Meridian prescription-syndrome system, Ni Haixia's Jingfang syndrome differentiation approach, and medical case works by famous Jingfang school practitioners throughout history. This article lists specific capabilities by domain, explains that syndrome differentiation conclusions and prescription plans are the respective outputs of each domain, provides a list of trigger words and usage methods, and points out that safety red line rules are not disclosed, dosage and medical case content basis require independent verification, and outputs cannot replace physician judgment. Installation commands and configuration fields are missing from existing materials; the official page fetch failed, and no speculations have been made in this article.
Read MoreMATLAB Automation: A Python-Based End-to-End Toolkit for Driving MATLAB
Introducing the community Skill "Matlab Automation": an end-to-end library of solutions for invoking or driving MATLAB via Python. It covers 7 distinct paths, including the official Engine API, command-line `matlab -batch`, Windows COM automation, Oct2Py (GNU Octave), MATLAB Production Server client, SMOP transpiler, and .mat/CSV file exchange. Additionally, it introduces Simulink programmatic control (covering programmatic modeling, parametrization, simulation data extraction, signal recording, batch parameter sweeps, code generation, and three invocation paths). The package contains 69 case studies verified locally on R2022a (10 basic + 20 functional + 15 domain-specific + 14 Simulink + 10 advanced/complex), as well as two folders for offline reuse: `references/` (solution comparisons, API quick references, data type mappings, and version compatibility adaptation by release) and `cases/` (orchestration scripts that generate one-click PASS/FAIL reports). The orchestration scripts automatically detect `D:/Program Files\MATLAB\*/bin\matlab.exe` and utilize byte-mode fault-tolerant decoding for GBK output on Chinese Windows systems. All MATLAB driver code is inlined within the Python scripts, and no .m source files are included.
Read MoreShort Story Editor: Organizing Fanqie Platform Review Standards into a Reusable Skill
Short Story Editor is a community-level content creation Skill published on SkillHub (Coordinate: @user_44ce364b/tomato-novel-editor, Version 1.0.0). Classified under "content-creation," it requires no separate API key application. It codifies the manuscript review experience of a senior Tomato Platform editor into a set of SKILL.md instructions. Using "Ming Ge" as the review persona, it employs a seven-level grading scale (A+ to C) and breaks down judgment across thirteen dimensions: title, opening, protagonist, pacing, hooks, writing style, market fit, AI-flavor, supporting female roles, villains, system elements, social realism, and the design of "golden fingers" (cheat codes) and thrilling satisfaction. Issues are categorized into four severity levels: critical bugs, severe, moderate, and minor. For critical bugs and severe issues, specific revision suggestions are provided, while moderate and minor issues are merely pointed out without detailed guidance. It is suitable for authors submitting to the Tomato platform, editors or studios conducting preliminary manuscript screening, and individuals wishing to evaluate outline structures before writing. It is important to note that this tool functions solely for review and does not write the content on the user's behalf; furthermore, there are no rewriting demonstrations for moderate or minor issues, and short text segments may be insufficient for judgment across most dimensions. As the pre-fetched data did not include official installation commands or client configuration snippets, this article does not provide command-line installation methods. It is recommended to refer to the documentation of your specific client and verify outputs using a small sample text first.
Read Moredsh-submodel：Official Subagent Default Delegation Policy Package
dsh-submodel is a community workflow plugin that does not create a new provider; it only reads existing routes from the profile, writes "when to do it yourself, when to delegate, and who to delegate to" into the system prompt, and registers delegate_task. Based on the GitHub README, this article explains the installation commands, the minimal cordis.patch.yml configuration, the guard behavior for Flash/non-Flash main and child agents, the boundary with the official subagent-control, as well as fail-fast log troubleshooting and the DSH 0.2.1-rc.x compatibility statement.
Read Moredsh-connect-teleagent: Connect local TeleAgent to the DeepSeek Harness model selector
Introducing the community plugin **dsh-connect-teleagent**: Given that the TeleAgent desktop client is already logged in on macOS, this plugin registers the Tianyi AI "Five-Tier" models into DeepSeek Harness. It provides an OpenAI-compatible local shim, true streaming, multi-turn conversations, tool pass-through, image recognition for specific models, and credit/quota display. The documentation covers three installation methods (install script, `dsh plugin`, and marketplace index), upstream `TELEAGENT_API_KEY` configuration, default kernel transmission, and `doctor` diagnostics. It also cautions users about protocol reverse-engineering, as well as account and permission risks.
Read MorePlain Comics Official: Translating Medical, Financial, and Tech Jargon into Layman's Terms and Visual Cartoons for the General Public
This article introduces the Agent Skill "Baihua Manga Guan" (Plain Language Comic Officer) (@user_06b9b4ad/skill-002, version 1.0.0, source: community, category: content-creation). It is designed to handle requests such as "explain in plain language," "create comics or illustrations," and "explain to patients, family members, the elderly, children, or non-specialist managers." By taking terminology, reports, or explanatory materials from fields such as healthcare, economics and finance, technology, and policy as input, it outputs plain-language Chinese explanations, structured visual diagrams, or original, mobile-optimized, single-column educational comics tailored to the specified audience. Based on the retrieved data, the article verifies its core capabilities: deterministic Chinese text bubbles, factual mapping (preserving numerical values, units, applicable conditions, uncertainties, and source citations), alerts for common misconceptions, and comprehension validation. It also clarifies high-risk boundaries: content related to healthcare and economics/finance requires source verification and expert review, and the output must not be construed as medical diagnosis or investment advice. As the provided materials do not include installation commands such as `git clone` or `npx/uvx`, nor do they detail configuration differences for Cursor, Codex CLI, or Claude Code, the article only provides the official API URL (https://api.skillhub.cn/user_06b9b4ad/skill-002) and the SkillHub page URL (https://www.skillhub.cn/skills/user_06b9b4ad/skill-002), without fabricating installation steps or configuration fields.
Read MoreDeveloping AI Assistants: Developer-Focused Skills
"Dev AI Assistant" is a development-related Skill on SkillHub, identified as @kunlungrowth/kg-dev60-021. Version 1.0.0, published by Kunlun Growth, categorized under dev-programming, sourced from enterprise, and tagged with requires_api_key=false. Invoked via trigger words such as "开发", "编程", "调试", and "dev", it is positioned as an AI assistant specializing in the development field to reduce the need for repeatedly providing background context. Based on official page materials, this article introduces its positioning, verifiable capabilities (trigger mechanism and domain scope), the general SKILL.md activation method, and typical trigger examples. It also points out that the official page was inaccessible during data scraping, and the materials did not provide installation commands or a granular feature list. It is recommended to verify the official address and latest instructions before installation.
Read Moredsh-retry-panel: Disconnect Retry and Multi-Layer Automatic Recovery in DSH Settings
The Amer-CN DSH plugin *dsh-auto-retry* (*dsh-retry-panel*) provides a "Connection Retry" card in the settings page, featuring four-tier request-level retries (including handling of QUOTA codes), max-tokens continuation within the same turn, turn-level auto-continuation with circuit breaking, turn compression with cold session recovery, and optional watchdog and backup model switching. This article introduces the bundle installation steps, main configuration items, compatible host versions, and usage boundaries. The plugin writes `retryPolicy` via a same-origin API and does not persist credentials.
Read MoreSkill Quality Check: Pre-release Health Check for Custom Skills
"Skill Quality Check" is a quality review Skill for community self-built skills, located at @user_f1d11e73/zfl-skill-quality-checker, version 1.11.49, published by user_f1d11e73, sourced from community, and categorized under dev-programming. It standardizes manual pre-publication checks into automated reviews, with three main capabilities: hallucination prevention rule compliance checking, skill directory generation, and quality scoring reports. It is applicable to quality reviews after skill creation and before publication. On the SkillHub page, it has a rating of 100000 and 1184 downloads, with the tag requires_api_key=false, indicating no additional API Key is needed for invocation. The article explains what it is, its core functions, installation and activation methods (based on the universal SKILL.md format, placed in the local skills directory for agent recognition), and reproducible usage based on the official description. It is important to note that the installation command and trigger parameters on the official page could not be verified during this capture, and relevant details in the article are marked as unverified. Please open the official address to confirm before use. This Skill is a review component rather than a runtime component and cannot replace manual testing of the actual running effects of skills.
Read More@altamkp/dsh-middle-click: Close Sessions and Sidebar Tabs via Middle Click in the DSH Web Interface
A DSH client plugin maintained by altamkp that allows archiving sessions or closing dock tabs via middle-click in the web sidebar, with gesture behavior consistent with Harness tabs. A dedicated settings page controls these two switches, with preferences stored in localStorage. Version 1.0.3 requires Harness 0.2.1-rc.2+ and a middle-click mouse. Installation is performed by declaring the `github:altamkp/dsh-middle-click bundle` in the profile package.json or via the plugin manager's "Install bundle". Extracted from the dsh-plugin collection v2.8.0; when migrating, ensure it is not loaded redundantly with the old collection. MIT License.
Read Moredsh-local-devkit: Completing the Local Ground Truth Layer for Harness
dsh-local-devkit supplements the local factual layer for DeepSeek Harness: `env_probe` and `health_probe` inspect the environment and functional outputs; `git_*` tools provide read-only and controlled commit capabilities; automatic checkpoints are created before writes, allowing rollback via `fs_restore`; `project_check` executes typecheck, lint, test, and build based on Node, Python, or Go project declarations. It uses pure built-in Node modules with no third-party dependencies. Installation requires running `node scripts/install.mjs` in the plugin directory and requires DSH >= 0.2.1-rc.3.
Read More