Tactical Coach for the Full Bidding Process: Breaking Down Bidding into a 7-Step Actionable Workflow
# Foreword The daily routine of a bidding specialist is often fragmented: finding projects relies on scrolling through announcements; decisions on whether to bid are based on gut feeling; reading tenders and breaking down lists are done in silos; proposal writing lacks scoring orientation; compliance checks are often skipped before submission; and there is no post-submission review. The Skill introduced below attempts to integrate these steps into a cohesive pipeline. # What is This "Bid Full-Process War Room Coach" is a Skill published by Kunlun Growth on SkillHub (located at @kunlungrowth/bidding-warroom-coach, version 1.0.0, source: enterprise, category: business-ops). Targeting bidding specialists, sales, and project teams, it links the process from "finding projects -> determining bidding feasibility -> reading tenders and breaking down lists -> writing proposals -> compliance review -> submission -> post-mortem" into an executable workflow...
Read MoreDon't-Stop: Setting Deadlines and Driving Continuous Iteration for DSH Agents
Dont-Stop (dsh-dont-stop) is a DSH workflow plugin maintained by TCY07. By setting a deadline via `/dont-stop`, it automatically prompts the agent to iterate and autonomously proceed to the next steps upon early completion. It supports extended retries for rate-limiting and network errors, time-limited questioning, timely task finalization, and optional cross-platform system sleep prevention. Based on the official README, this article introduces the installation commands, slash command usage, and configuration options, making it suitable for long-duration, unattended task scenarios.
Read MoreWeb Novel AI Polish Rewriter · Four-Layer Deep Clean: Eradicating AI Traces Layer-by-Layer from Word Frequencies to Structure
"Web Novel AI-Tone Remover · Four-Layer Deep Clean Edition" (`@user_d5cdcaa0/novel-deai-rewriter`) is a paid content creation Skill published by user_d5cdcaa0 on SkillHub (Version 1.1.0). The core rewriting engine operates in the cloud, processing content through a four-step sequence: statistical fingerprint scanning, structural diagnosis, text-level blacklist replacement, and commercial packaging checks. This package contains only payment orchestration scripts and invocation instructions, excluding the rewriting rule files. Upon invocation, it returns comparative data on four metrics—blacklist word frequency, sentence length variance, paragraph opening patterns, and metaphor density—to evaluate the quality of the rewriting. This Skill is triggered when users request to "remove AI tone," "make it sound human," "Tomato De-AI," or submit AI-generated novels for rewriting. It is priced at ¥0.3 per call, settled via Alipay AI Pay-As-You-Go, and requires no API Key.
Read MoreRewind: Snapshot files edited in WorkBuddy with one-click rollback
Rewind is a community Skill on SkillHub (@user_09b74654/rewind, version 1.1.3, categorized under office-efficiency) that creates snapshots and backups for files edited in WorkBuddy. It supports one-click rollback to any historical version for files including Word, PPT, Excel, PDF, and notes, and is compatible with both macOS and Windows. The Skill can be triggered via natural language commands such as "rollback", "undo", "restore", "rewind", and "revert", and can also proactively create snapshots prior to extensive modifications. The source materials do not include installation commands, configuration fields, or reproducible command examples, and none have been fabricated in this text.
Read Moredsh-day-memory: DSH Level 3 "Daily Episode" Long-term Memory
Maintained by Tarolcraze, dsh-day-memory provides DeepSeek Harness with a day-based, three-level episodic memory: Level 1 index retrieval, Level 2 summarization, and Level 3 raw text and image copies, along with next-day sub-agent distillation. Resident mode requires panel confirmation. Based on the README, this article explains its differences from the semantic memory plugin, installation (desktop bundle / Web link), configuration options, and the usage of memory_* tools, and lists known limitations.
Read Moredsh-plugin-feishu: Running DSH Sessions via Persistent Connections in Feishu
dsh-plugin-feishu is a DSH client plugin maintained by suveng that sends and receives messages via Feishu/Lark WebSocket long connections, without requiring a standalone bridge process or public webhook callbacks. It supports private and group chat topics, conversation persistence, streaming cards, execution process panels, slash commands, and permission whitelisting. Install with `dsh plugin --profile web add`, requiring Node.js 22+ and DSH versions from 0.2.1-rc.2 up to below 0.3.0. Configuration can be done by scanning a QR code in the sidebar or filling in App credentials. Note that the plugin runs with the dsh process permissions, and for personal use it is recommended to restrict allowedOpenIds.
Read MoreIELTS/TOEFL Score Improvement Engine: Delegating Four-Section Strategies to Agent Skills
This article introduces a community-sourced IELTS/TOEFL score-boosting engine Skill: version 1.2.3, source community, category education, publisher user_8ee65e73, SkillHub rating 100000, downloads 1099, installs 0, tag requires_api_key=false. It organizes methodology across four modules—Listening, Reading, Writing, and Speaking—and provides past exam and prediction banks, a writing correction framework, speaking practice coaching, and a 30-day sprint plan. It defines a mandatory trigger routing: invoke immediately when users mention IELTS/TOEFL/IELTS prep, or paste original text, links, or screenshots; do not ask clarifying questions first, and do not substitute with generic capabilities. The official page does not provide a command-line installation method, nor does it list differences among Cursor, Codex CLI, and Claude Code; therefore, this article only provides the SkillHub page and API entry, without speculating installation commands. Reproducible use cases include triggering by subject keywords, determining the subject by material type when materials are provided directly, following the 30-day plan for pre-exam sprint, iterating writing using the correction framework, and using the speaking practice workflow. Notes: the past exam predictions are highly time-sensitive; triggering depends on routing configuration; with 0 installs, it is recommended to validate it yourself; the official score-uplift claims are statements by the publisher and have not been independently verified.
Read MoreRecreating Viral Stickers: Deconstructing Trendy Structures to Recompose Unique Designs
This article introduces the community Skill "Replicate Viral Images" (**@indiv-yunlong/tietu-refresh**, version 1.0.0, categorized under content-creation). It deconstructs existing viral images into visual skeletons and viral mechanics, providing three non-homogeneous refurbishment concepts. It only generates new images after a concept is selected, accompanied by caption writing directions and hashtag suggestions. Applicable to any niche, it supports trigger keywords such as replicating viral images, refreshing images, creating benchmark images, refreshing old images, analyzing images, refreshing this image, homogenization, and creating similar yet distinct images. No independent installation command is provided officially; it is used via the SkillHub page following the universal SKILL.md format. **requires_api_key=false** indicates that the Skill itself requires no extra key, though generating new images relies on the host environment's multimodal image capabilities. The article also outlines the typical workflow and four key precautions regarding copyright, output quality, sample overfitting, and timeliness.
Read MoreManagement Report PPT: Converting Finalized Excel, Word, and Markdown into Presentation Materials
This article introduces the SkillHub community skill "Management Report PPT" (@user_e9d33879/management-report-ppt, version 1.1.0, categorized under office-efficiency). Based on finalized Excel, Word, and Markdown materials, it produces management-facing presentation decks through a five-stage pipeline: material extraction, narrative restructuring, style reuse, generation, and verification. It is suitable for scenarios such as "turning this report into a presentation," "creating materials for management," or "generating slides based on Excel and Word." However, it is not intended for creating thematic presentations from scratch; for such requirements, `tencent-pptx` should be used. The article provides the skill's identifier, capability layering, and reproducible typical use cases. It also notes that the metadata indicates `requires_api_key=false`, and as no official CLI installation commands or client-specific configuration fields are provided, no installation scripts or configuration items are fabricated. Furthermore, it emphasizes that source materials must be finalized, and key metrics and data calibers still require verification after generation.
Read MoreInvoking Codex and Grok CLI via TokenRouter within DeepSeek Harness
dsh-tr-bridge is an MIT-licensed open-source DeepSeek Harness Host plugin (v0.1.0-rc.1 preview) maintained by superqin7. It calls the user's existing tr codex and tr grok locally via tr_bridge(agent, task, dryRun) and does not handle TokenRouter API Keys. The article details installation paths, dryRun and actual invocation examples, capabilities, and limitations. It emphasizes that end-to-end functionality on Windows DSH 0.2.1-rc.5 has not yet been verified, making it suitable for cautious trial by Windows users who have already successfully configured the TR pipeline.
Read MorePodcast Downloader: Save Xiaoyuzhou Audio and Show Notes Locally
Podcast Downloader (@zcwl/podcast-downloader) is a Xiaoyuzhou (Little Universe) download skill released by Zhichuang Weilai, categorized under dev-programming, version 1.0.1. It can download podcast audio and Show Notes based on the show link and automatically convert the audio to MP3 format, with the tag requires_api_key=false, so no additional API key is required. This article introduces its capabilities and limits, installation entry point, and applicable scenarios, and notes that the installation command and differences across hosts are not provided in the materials and should be confirmed via the official page.
Read Moredsh-sidebar-panel-collapse: One-click Collapse of Sidebar Panel List, Reclaiming Workspace Height for the Conversation List
The DSH Web client plugin dsh-sidebar-panel-collapse maintained by Sunday530 adds a chevron collapse button under “New Session.” With one click, it hides entire panel entries such as the MUV table, plugin marketplace, and skills, freeing up vertical space for the workspace list. The state is stored in localStorage; the sidebar rail and its width remain unaffected. Install: dsh plugin --profile web add dsh-sidebar-panel-collapse. After restarting DSH, it can be toggled in the plugin marketplace. MIT licensed. The README test version is 0.2.1-rc.5.
Read MoreDACW: A Defensive Programming Workflow with Writers, Runners, Reviewers, and Human Gatekeepers
DACW (@stmcw1/dacw) is a DSH Host workflow plugin: writers produce JSON change sets, deterministic runners execute whitelist tests, independent reviewers issue JSON verdicts, all paths halt at NEED_HUMAN, and nothing is ever auto-merged. Supports subagent zero-tool isolation, workspace gates, optional Docker sandboxing, and configurable writer/reviewer models. Installation requires DSH ≥ 0.2.1-rc.5; install the tarball via install_bundle and enable dacw in the profile; the typical entry point is the dacw_run tool.
Read MoreRetail Channel & R Analysis: Store Count Tells Me Channel Judgment
This is a data-analysis Skill for the retail vertical, published and maintained by community developer user_c8cbc810, version 1.0.0, source: community, category: data-analysis. It converts store counts, store format composition, city distribution, and competitor coverage of chain brands—such as convenience stores, supermarkets, pharmacies, and beauty retail—into analysis outputs that can be used directly to identify core markets and growth opportunities. It provides four capability areas: store network analysis, store format classification analysis, regional coverage analysis, and brand comparison analysis. The article describes its positioning and these four capabilities, provides the local directory structure for SKILL.md and API key configuration requirements, lists reproducible use cases for city rankings, breakdowns by store format, and brand coverage comparisons, and notes caveats such as requiring an API key, data being sourced from external APIs, coverage being constrained by data sources, and community-sourced content may change with updates.
Read MoreDrafting Audit Records: The Deterministic Output of On-site Audit Record Forms
Introducing the "Audit Record Writing" Skill on SkillHub (ID: @user_7a8ab90c/mm202608001, Version 1.0.5, Category: office-efficiency, Source: community, No API Key required). It generates "On-site Audit Checklists" for management system certification on-site audits. It allocates tasks based on audit plan clauses and system documentation, internally structuring data via a 4-field dictionary (Basis/Action/Fact/Conclusion) to output pure narrative, objective records without tags, arranged in ascending order by clause number. It features a unified fixed 4-column table format (fixed column widths, SimSun font size 5, objective record column width ≥60%, pure black and white with no background color). It supports the QES three systems (Q:/E:/S:, GB/T 19001/24001/45001), FSMS/HACCP (F:/H:, ISO 22000:2018), as well as ISO 28000/GB/T 35770/RB-T 089. It performs Y/N determination for each item and issues non-conformities. Word documents must be generated via scripts/gen_optimized.py; after the format is locked, a format self-check gate is executed (no soft line breaks/no blank lines/first-line indent=21/no background color) to ensure consistency. Upon completion, data is written to the "Audit Data Package" for reuse in audit report writing. It includes 7 official trigger words; the AI generates the draft, while final audit conclusions are the responsibility of the human auditor. Current platform data: Rating 100000, Downloads 1257, Installs 0. The public version contains only the SKILL.md methodology; scripts and industry templates must be obtained separately from the author.
Read Moredsh-quote-annotation: Submitting Conversation Excerpts and Annotations as Structured References to the Model
The DSH plugin **dsh-quote-annotation**, maintained by stickyfinger7, lets you select conversation text in Chat and optionally add an annotation. The Host then injects a hidden `conversation-quote` reference along with your next message at the agent/pre-step, keeping the composer clean and never exposing raw JSON. It supports pill display, jump-back-to-source, and inline markers in assistant replies, and sent quotes can be restored from the Session log. Compatible with DSH 0.2.x Web/Desktop (Node.js >=22.19.0). Install with the official command `dsh plugin add github:stickyfinger7/dsh-quote-annotation`. MIT licensed.
Read MoreNeuroresearch: A Full-Process Biomedical Research Assistant Skill
neuroresearch is a specialized taxonomy Agent Skill published by community developer user_aabe55f8 (located at @user_aabe55f8/neuroresearch), currently at version 3.1.20. Positioned as an assistant for the entire biomedical research workflow, it covers seven stages: literature review, bioinformatics analysis, experimental design, statistical analysis, publication-quality plotting, paper writing polishing, peer review simulation, citation verification, and grant proposal writing. Literature reviews default to the PRISMA flow, while hypothesis generation utilizes 11 innovative operators to provide a "kill order." Bioinformatics tasks utilize R, statistical analysis employs the Python engine, and scripts are standardized with version numbers and random seeds. Deliverables default to Word format. It does not require separate API Key configuration, activates via trigger words, and explicitly excludes clinical diagnosis and treatment recommendations.
Read Moredsh-friend: DSH Desktop Companion and Optional Screen Observation
steady-sakura's DSH client plugin, dsh-friend (Little Whale Desktop Companion), provides chat bubbles, 106 dsh-pet actions, a built-in Chat Completions API, authorized window/screen sharing, and local memory retrieval within an Electron desktop pet environment. The installation command is `dsh plugin --profile web add github:steady-sakura/dsh-friend`; it requires Node.js 22.12+ and recommends DSH 0.2.0-rc.2. Sending memory or screen data to the API requires separate opt-in consent. For uninstallation and backup procedures, please refer to the official README.
Read Moredsh-ascii-scenes: Powering DeepSeek Harness with Eight Interactive ASCII Scenes
dsh-ascii-scenes (Srelka, v0.5.0) provides DeepSeek Harness with eight interactive ASCII environment themes, covering both sidebar and main workspace. It supports scene switching, close animations, and native appearance restoration. The plugin does not read session data, calls no models, and makes no network requests. Preferences are stored in localStorage. Installation command: dsh plugin --profile desktop/web add github:Srelka/dsh-ascii-scenes#v0.5.0. Compatible with dsh >=0.2.1-rc.3 <0.3.0-0. MIT license.
Read Moredsh-gpt-image-playground: Hosting GPT Image Playground on DSH
dsh-gpt-image-playground, maintained by southcul, vendors CookSleep's officially built GPT Image Playground v0.8.4 into DeepSeek Harness. It provides an image-generation workbench in the left column, with same-origin hosting and an optional API proxy to handle CORS. The plugin keeps a persistent iframe through an overlay to avoid unloading in-progress generations when switching the main panel. Installation is recommended via the npm package name `dsh-gpt-image-playground` or by running `dsh plugin --profile desktop add`. On first use, configure the API URL and Key in the settings; a proxy at `/plugins/.../proxy` can be used to address cross-origin issues. It fully inherits features including the gallery and Agent, reference images and masks, multi-provider support, and more. MIT license.
Read MoreFamily Finance Manager: Transforming Raw Bills into Analyzable Income and Expense Data
The family finance manager is @user_7e31ec9f/family-finance-manager-pro, Community source, Professional category, version 3.2.0. It converges scattered billing entry points into a repeatable pipeline: collecting bills from IMAP email inboxes, decrypting and decompressing encrypted archives, cleaning and importing WeChat/Alipay statement exports into the database, computing refunds, using a V3 four-layer self-learning pipeline for automatic transaction categorization, and customizable spending categories. It bundles 9 composable sub-skills (under the skills/ subdirectory), and email app credentials and archive passwords are stored locally in encrypted form. An API key is required. The official documentation does not provide installation commands or configuration fields and is described accurately as-is, pointing to the official repository for details. The article provides a faithful description based on the official sources, with no fabricated cases or data.
Read Moredsh-memory: Implementing Human-like Long-term Memory for DSH Agents
dsh-memory is a local memory plugin (v1.1.0, MIT license) developed by Soulmate793 for DeepSeek Harness. It replaces monotonically growing vector/conversation stacks with Ebbinghaus forgetting curve, retrieval strengthening, sleep-based consolidation, and cue-based recall, and provides seven memory partitions along with an Ed25519 hash-chain audit. Installation is enabled via a DSH bundle or by enabling cordis.patch.yml in your profile. By default, data is stored in ~/.dsh/qq-memory. Based on the official README, this article covers installation, configuration, usage of the memory_remember and memory_recall tools, and related notes.
Read Moreopen-pptd: Generate Editable Native PPTX
open-pptd is a Skill designed for large model Agents, located at @user_cfa12f26/open-pptd, version 1.5.0, sourced from community, and categorized as office-efficiency. It uses a proprietary PPTX engine to directly generate native .pptx files, where text, charts, and formulas are editable in PowerPoint — not image collages, PDFs, web page saves, or SVG/HTML-to-PPT conversions. Its capabilities cover creating, editing, copying, reading, and exporting presentations, and it requires no API Key. This article introduces its positioning, core features, installation entry point, typical workflows, and applicable scenarios, and notes that the documentation does not provide specific installation commands or reproducible examples, so users should refer to the official page for details.
Read MoreHandwritten Poster Content Assistant: Generating Headers, Section Text, and Layout Suggestions by Theme
# Introduction When children bring home a handwritten poster assignment, parents often struggle not with the drawing itself, but with the lack of ideas for content: the theme is set, but they still need to brainstorm the masthead title, draft the body text for each section, select borders and decorative elements, and figure out a layout that doesn't look messy. Finding scattered materials online or improvising is time-consuming. The "Skill" introduced below hands this content preparation work over to an AI model. It generates the title, section text, quotes, decorative border suggestions, and layout advice based on a single theme input, allowing children to simply copy the text and follow the design while drawing. # What is this? **One-sentence description:** A content generation Skill tailored for primary school handwritten posters. It takes a theme as input and outputs a ready-to-use content list that can be written and drawn immediately. - **Name:** Handwritten Poster Content Assistant - **ID:** @user_140a98d6/shouchaobao-zhushou -
Read Moredsh-deep-reading-wl: OCR and Reading Skeleton Pipeline for Scanned Textbooks
dsh-deep-reading-wl is a DSH workflow plugin designed for scanned PDF textbooks lacking a text layer. Through a four-step process—ocr (RapidOCR+OpenVINO), scan (chapter skeleton and table positions), map (chapter-based reading focus), and digest (compressing table fragments)—it produces a skeleton map for pre-reading insights rather than reading the entire book on your behalf. Installation requires junctions under the profile, package.json links, and cordis.patch.yml insertions; Python 3.10+ and Node 22+ are required. MIT 0.2.0.
Read More