build-wechat-miniprogram: WeChat Mini Program Delivery Skill from Setup to Content Security Auditing
build-wechat-miniprogram is a WeChat native mini program delivery skill released by the SkillHub community. Identified as @user_c25e780d/build-wechat-miniprogram (Version 1.0.1) and categorized under dev-programming, it requires no API Key. It decomposes mini program development into three core capabilities: At the page and engineering level, it covers WXML/WXSS, native components, TDesign MiniProgram, usingComponents, npm builds, WeChat Developer Tools, and real-device debugging; at the runtime level, it addresses differences between Skyline and WebView, CloudBase cloud functions, WeChat login, permissions, and privacy declarations; at the content safety level, it covers the moderation of avatars, images, audio, nicknames, and comments, involving security.mediaCheckAsync and security.msgSecCheck, as well as the troubleshooting of security mode HTTPS callbacks, pending states, review processes, callback failures, and upload failures. Additionally, it includes publishing verification, adaptation, elderly-friendly features, and accessibility. The article provides official coordinates, addresses, and typical usage scenarios based on trigger keywords. It also specifies that this Skill is exclusively targeted at WeChat native mini programs and does not cover cross-platform frameworks. Users are advised to verify interface parameters against official documentation before use.
Read Moredsh-voice: Text-to-Speech, Dictation, and Multilingual Voice Routing for DSH Web UI
dsh-voice (@lucagiftzek/dsh-voice 2.0.0) delivers speech output for assistant replies, composer dictation, supports configuration for 23 categories of TTS/STT providers under Settings > Voice, language-based routing, sentence and word highlighting, and the TTS Reader sidebar for the DeepSeek Harness Web UI. Supports npm or GitHub installation and requires restarting the DSH web service. Optional dsh-better-sidebar. Designed for DSH 0.2.x/0.1.5. Released under the MIT License and maintained by lucagiftzek.
Read Moresearch-baidusearch: WeCom CLI All-in-one Suite Covering Contacts, Messages, Documents, Schedules, Meetings, and To-Dos
This article introduces `search-baidusearch` (Coordinate: @user_33aa5c8b/search-baid, v1.0.0, Community Source, Office-efficiency Category), which is published on SkillHub. The tool consolidates commonly used WeCom (Enterprise WeChat) operations into CLI capabilities that is directly callable by Agent, covering six major business domains: contacts, messages, documents, schedules, meetings, and tasks. The specific functionalities include: searching contacts by name or alias; sending and receiving text, images, files, voice, and video messages; creating, reading, and editing documents in Markdown; managing the structure and records of smart sheets; creating intelligent documents with export-to-Markdown capability; creating, modifying, and canceling schedules, plus checking busy/free status; booking and managing meetings; and creating, tracking, and assigning to-do tasks. The official description indicates that this tool should be triggered in message, schedule, and task-related scenarios, even if the user does not explicitly mention WeCom. The metadata marked as `requires_api_key=false`. Installation commands and reproducible invocation examples are not provided, so this article honestly points out the facts accordingly and suggests referring to the official page for accurate information.
Read MoreAladdin AI Weekly Report Writing: Automatically Generate Weekly Reports from Activity Data With Skills
Aladdin-Weekly Intelligent Writer is a community-released Office productivity skill. Its core is automatic recall: after user consent, it collects work activity signs from WeCom, DingTalk, and Feishu, such as messages, schedules, to-do items, and document edits. It clusters data by project into structured weekly, daily, and monthly reports, supports exporting in multiple formats (Tencent Docs, WeCom, DingTalk, Feishu, local Word, and Markdown), and creates a personal work archive for cross-period comparison. It compresses 30–90 minutes of manual organization into a ~2-minute delivery. Unlike writing assistants that only polish writing styles, it automatically gathers source material instead of relying on user-provided texts. During the writing of this document, the official page failed to be fetched, so installation commands, configuration fields, and usage conventions could not be verified; this article explicitly flags that limitation honestly.
Read MorePackaging Generation: A Skill for Generating Packaging Design from Scratch
This article introduces the community skill Packaging Generation, a packaging design generation capability with built-in design standards, a template system, and content generation features. It supports trigger words such as packaging, create packaging, packaging design, product packaging, and gift packaging, and requires no API key. Currently at version 1.0.2, it originates from the community and is categorized under design-media. The article outlines its positioning, core features, installation entry point, and applicable scenarios, and notes that the official page could not be retrieved and that the installation command should follow the page.
Read Moredsh-subagent-tiering: Subagents Route to Parent Model's Low-Cost Same-Tier by Default
When the subagent does not specify the provider, model, or reasoning_effort, the DSH plugin @lucagiftzek/dsh-subagent-tiering routes sub-sessions to cost-effective, equivalent-tier models—such as Sonnet, Qwen flash, and DeepSeek flash—based on built-in rules. Explicit parameters passed by the main agent take precedence. It integrates via tools/pre-execute and cordis.patch, requiring the subagent-model-selection and ptc's tool-subagent to be enabled. To install, run `dsh plugin --profile web add @lucagiftzek/dsh-subagent-tiering` and restart dsh web. Node ≥22.18, MIT 0.1.0.
Read MoreTax Compliance for Labor Service Companies: Translating Common Practices Regarding Wages, Social Security, and Cost Invoices into Contractual Clauses
"Tax Practice Guide for Labor and Construction Companies" covers five core capabilities: compliant payroll distribution, risks related to transfers from corporate to personal accounts, handling of cost invoices, the relationship between social security and wages, and optimization of corporate income tax for small and micro enterprises. Published by user_8dbbd174, Community-sourced, Category: Professional, Version 1.0.2, No API Key required. The article explains the problems it addresses, how to install and enable it, typical example queries, and limitations as a community knowledge pack regarding regional compliance standards and timeliness.
Read MoreTask Planning Manager: Transforming Scattered To-Dos into Priorities, Time Blocks, and Next Actions
The Task Planning Steward is an Agent Skill based on the universal SKILL.md format, with identifier @user_b4da0893/proactive-task-manager-cn. It is published by user_b4da0893, sourced from the community, versioned at 1.0.1, categorized under ai-agent, and tagged with requires_api_key=false. It consolidates scattered to-dos into priorities, time blocks, reflection prompts, and next-step actions; it does not generate code, nor does it sync with calendars. The article outlines its positioning, four core capabilities, installation and activation methods (the official documentation does not provide installation commands; it only states that it should be placed into the skills directory as a SKILL.md-based skill, and no API key is required), use cases and limitations, and points out that the official page lacks examples and notes on directory differences across the three client apps. It suggests validating it with a minimal task set before deciding whether to integrate it into a daily workflow.
Read MoreLibrary Assistant: Using the Controlled Browser as an MCP Relay Node
Library Assistant (@yahuixun/library-assistant, developed by Shenzhen Yahuixun, version 1.0.5, categorized as an AI agent) integrates Control Browser Orchestrator as a headed browser node into MCP Relay. The article introduces its capabilities such as relay pairing, authenticated worker connections, secure browser operations with manual login, local MCP fallback, page navigation and screenshots, and extended scraping via configured extensions. It also provides reproducible steps for enabling, pairing, navigating, and capturing screenshots via the official endpoint, along with considerations including API keys and headed display environments.
Read Moredsh-request-image-cap: Send images to model with pixel limits based on side length
DSH's default image gate is area-controlled, with long edges up to 8192 px; full-page screenshots may be rejected by the 2000 px per-side rule during normal short conversations and multi-image Anthropic requests. dsh-request-image-cap clamps each image to a configurable per-side cap (default 2000) at the readImageRequest seam, without altering storage, and can retroactively fix historical sessions. Based on the official README, this article covers installation commands, configuration options, the behavior formula, and environment requirements (Node ≥22.18, DSH 0.1.5/0.2.1), and reminds users that the plugin runs with the same privileges as the dsh process, so the source code should be reviewed before installation.
Read MoreExpert Guide to Corporate Litigation Cases: The Framework of Causes of Action, Elements, and Evidence in Chinese Civil Litigation
This is a structured analysis Skill designed for Chinese civil litigation, located at @user_3fc4552d/company-litigation-case-guide-for-china. It is Version 1.0.1, community-sourced, categorized as professional, and requires no API Key. Its capabilities cover a four-step process: identification of the cause of action, analysis of the basis for claims, decomposition of facts constituting legal elements, and generation of an evidence list. The output is organized based on a progressive relationship where the cause of action determines the norm, the norm determines the elements, and the elements determine the evidence. The article introduces the Skill's positioning and four core functions. It notes that the Skill's fetched data lacks installation commands and examples; therefore, instead of providing installation steps, it describes the location, version, and tag fields for verification. It suggests enabling the Skill in the client following the general format of SKILL.md and recommends verifying the loading method for specific clients on the official page. The "Typical Usage" section avoids fabricating examples and instead provides the invocation format and engineering suggestions: input parties, litigation claims, facts, and existing evidence to retrieve output in four segments. It also advises verifying the stability of the cause of action classification and checking for any autonomously expanded bases for claims against real cases. Finally, it outlines the applicable scenarios for preliminary structuring of corporate disputes and the automation pipeline of element-fact-evidence mapping tables. It reminds users that the Skill is limited to Chinese civil litigation; results depend on the completeness of input; it does not judge the probative value of evidence; and as a community source with zero installations, it requires independent verification. The official address and SkillHub page are attached.
Read Moredsh-plugin-updates: Checks for updates by installation source and only prompts for those loadable by the current DSH
lucagiftzek's DSH web plugin @lucagiftzek/dsh-plugin-updates (0.1.2, MIT) does nothing but update notifications: it distinguishes sources such as npm, GitHub, and link by their package.json specifiers, avoiding false alerts caused by identical names; it also performs compatibility gating against the running DSH core, and does not install anything on its behalf. It provides an Updates sidebar, Settings, a read-only agent tool, and a doctor API. Installation requires Node ≥22.5 and DSH 0.1.5-rc.1 through 0.3.0 (not bundled); see the official README for commands.
Read MoreGitHub Operations: Direct Repository Manipulation by Agents
GitHub Operations is a Skill that encapsulates GitHub repository management operations into callable capabilities, with coordinates @indiv-ai/github-operations, version 1.5.0, categorized as dev-programming from a community source. It provides Create, Read, Update, Delete (CRUD) capabilities for four types of objects: repositories, branches, Pull Requests (PRs), and Issues, explicitly excluding repository cloning. It is marked as requires_api_key=true, so the corresponding API Key must be prepared before activation. With a rating of 100000, 896 downloads, and 0 installations in the data, the actual number of adopters is limited, meaning its reliability requires independent verification. This article introduces its positioning, core functions, installation and activation precautions, typical invocation examples, and applicable scenarios according to its capability scope. It also points out potential pitfalls such as the community source, the irreversibility of write operations, and the need to confirm permission scopes. The official address is https://api.skillhub.cn/indiv-ai/github-operations.
Read MoreAudio Mixing Tool: Help Align Loudness and Blend Music with Sonic Effects
A Community Skill for game audio, film/TV short videos, and podcast post-production, located at @user_3a752629/audio-mixing, identifier: sound-mixing-tool, v1.0.0, category: design-media. The core lies in precise music/sound effect loudness alignment and fusion mixing workflow: using ffmpeg's two-stage loudnorm to compress LUFS deviations of multiple material sets to <0.10, followed by minimal intervention fusion and export of independent tracks ready for engine integration. It covers four scenarios: standard track-by-track matching, merged reference matching, Idle sound effect matching, and music as atmosphere matching. It strictly prohibits ducking/sidechain techniques, relying solely on two-stage matching and filter chains to ensure clarity of both music and sound effects. Trigger words include mixing, loudness matching, LUFS alignment, MIX version, music/sound effect separate tracks, fusion, no clashes, can't cover. Typical input is a material directory with multi-version subfolders, requiring matching of music/sound effect loudness from Set_X to the reference Set_Y, followed by fusion mixing, and final export of the mix versions plus separate tracks of music and sound effects, which can be used within engines such as Wwise, FMOD, etc. Note, official materials do not provide a copy-and-paste installation command, nor do they distinguish loading methods between Cursor / Codex CLI / Claude Code; users should verify this at the official address.
Read Moredsh-phone: Make and Receive Calls and Send/Receive SMS in the DSH Web Sidebar
dsh-phone (@lucagiftzek/dsh-phone) is a client plugin for DeepSeek Harness Web that provides a Phone tab in dsh-better-sidebar. It supports Twilio SMS/MMS, browser-based answering and DTMF, Live/Voicemail/DND policies, server-side voicemail and transcription, as well as phone_* Agent tools. Calls default to "Via server" and go through wss via your own server. Requires Node 22.5+, phone-hookd, and a public HTTPS webhook. Installation: dsh plugin --profile web add @lucagiftzek/dsh-phone.
Read MoreCommunication Integration (bluebubbles): Enabling Agents to Read and Write iMessage
Communication Integration (bluebubbles) is a Skill in the dev-programming category from a community source on SkillHub, located at coordinates @indiv-seafish/bluebubbles (Version 1.0.0, Publisher: user_c7f826a3). It caters to communication integration needs within the messaging and social domains. It is enabled when users request sending/receiving iMessages or reading history logs, with trigger words including bluebubbles, imsg, and wacli. It explicitly does not respond to requests unrelated to communication integration. Its capabilities include sending and receiving iMessage messages and reading history logs. It is marked as {requires_api_key=true} with a Risk Level of L3, and the documentation explicitly states: "High risk, authorization required with caution." The article introduces the pain points it solves, verified core functions, official address, and coordinate information. It truthfully notes that the retrieved documentation does not provide specific installation commands, directories, or configuration fields, nor does it explain differences regarding Cursor / Codex CLI / Claude Code, while advising users to verify on the official page. The article also lists typical trigger conditions, applicable scenarios, and precautions such as L3 risk, API keys, community source, and missing failure behavior and rate limit information, concluding with the official address.
Read MoreAnalysis of Guitar Chord Progressions and Correcting Listening Errors: Harmonic Analysis of Pop Songs and Diagnosing Misheard Chords
This article introduces a community skill: "Guitar Chord Progression Analysis & Ear-Training Error Correction" (@user_5314b4d6/guitar-chord-analysis-bobo, v1.0.0, category: education, requires_api_key=false). This skill addresses two common bottlenecks encountered when playing guitar: inability to interpret the chord progressions in popular songs, and the difficulty of identifying the root cause when mishearing chords during transcription by ear. It delivers the following core capabilities: analyzing the modality, functional progressions, and cadence patterns within popular song chord progressions; comparing the discrepancies between the chords the user heard and the correct chords, and diagnosing the underlying causes of misperception; and providing targeted training recommendations. Typical trigger scenarios include analyzing chord progressions, comparing misheard chords, exploring why certain chord progressions sound pleasing, and learning to play and sing without sheet music. The article also provides the skill's coordinates and positioning, reproducible example queries, and applicable notes, while clarifying that official documentation does not provide a standalone installation command—installation must be performed via the client's Skill loading mechanism or the SkillHub page.
Read Moredsh-orchestrator: Unified Orchestration of Multi-Agent and Out-of-Conversation Tasks in DSH
dsh-orchestrator (@lucagiftzek/dsh-orchestrator) adds an Orchestrate page and 14 orch_* tools to the DeepSeek Harness web side, supporting spawn, prompt, fork, and fan-out agent chats, as well as four types of traceable tasks: command, scrape, http, and agent. Runners cover local, SSH, and remote tool server; Providers display HPC or HTTP JSON status; the machine API enables automation and calls from other hosts. Requires DSH >= 0.1.1-rc.1, Node >= 22.5, and dsh-better-sidebar; restart dsh-web after installation.
Read Moredsh-note-keeper: Running an Obsidian-compatible local vault in the DSH sidebar
Note Keeper (@lucagiftzek/dsh-note-keeper) is a DSH Web plugin that provides an Obsidian-compatible local Markdown vault in the sidebar. It features wikilinks, tags, full-text search, backlinks, and a graph view, while supporting voice transcription (via dsh-voice), a retro drawing board with OCR, client-side zero-knowledge encryption, and various import/synchronization methods. The installation command is `dsh plugin --profile web add @lucagiftzek/dsh-note-keeper` (Linux x64 includes a pre-compiled daemon). Additionally, it exposes 12 `notes_*` tools, allowing in-session models to read and write notes within guardrails. Maintained by lucagiftzek, licensed under MIT, with the current npm version at 0.3.0.
Read MoreIntroduction to the SMZDM Deal Aggregation Skill
This article introduces the "SMZDM Deal Aggregation" Skill on SkillHub: published by RealQ (Version 1.0.0), categorized under life-service, sourced from enterprise, and tagged with `requires_api_key=false`, it is designed to aggregate deal leads from SMZDM, such as extreme bargains, BUG prices, and historical lows. As the official page does not provide verifiable installation commands or configuration fields, the article recommends enabling the Skill via the official page and provides a request template. It also cautions that deal information is highly time-sensitive and results must be verified via the source links; additionally, due to conflicting data regarding ratings and download counts, specific figures are not cited in the text.
Read MoreChatCut Alternative | Filler B-roll for Voice-overs & Picture-in-Picture Matching: Skill Introduction
This article introduces the "ChatCut Substitute | Voiceover B-Roll Coverage and Picture-in-Picture Matching" skill on SkillHub, located at @infimind/aihive-chatcut-broll-coverage, version 1.0.0, categorized under design-media, published by Jirui Technology - AI Expert, with the tag requires_api_key=false. The skill targets ChatCut blank shot, B-roll matching, and voiceover picture-in-picture asset needs, following the workflow of AI-HIVE login, MCP connection, and capability check, then reading user-authorized materials and delivering four artifacts: broll-coverage.csv, missing-shots.json, overlay-review.md, and an acceptance record. It is not an official third-party plugin; when media tools are missing, it only delivers a solution without producing final footage. Official materials do not provide installation commands or reproducible examples, so this article only confirms the coordinate, version, and category, without speculating on integration methods for Cursor, Codex CLI, or Claude Code, and outlines considerations such as authorized materials, capability checks, and the absence of public usage feedback.
Read MoreBrowser Use Agentic Browser Control: A Controllable Browser for AI Agents
This article introduces the Browser Use Agentic Browser Control on SkillHub (@user_3c6cb52e/browser-use-agentic-browser-control, version 1.0.0, source: community, category: ai-agent, tag: requires_api_key=false). It packages the open-source browser automation framework Browser Use into an Agent Skill. By integrating the Python SDK, browser orchestration, and model integration, it transforms websites into controllable interfaces, enabling AI agents to navigate real browsers, extract data, and execute multi-step workflows. The article first outlines the pain points it addresses (such as the failure of single-step HTTP requests or static scraping instructions on real-world websites). It then lists verified core capabilities, official pages and API endpoints, and typical usage patterns (note that the provided material lacks reproducible examples and only describes scenarios from the official introduction). Finally, it covers applicable scenarios and precautions, including runtime dependencies, the demand for model planning capabilities in long workflows, and the recommendation for small-scale verification given its current install count of zero.
Read MoreDSH Idle Watchdog: Idle Session Urging and External Prompt API
dsh-idle-watchdog is a DSH web plugin (networking tool) maintained by lucagiftzek. For opt-in sessions, it automatically sends follow-ups when the agent is idle and exceeds the silent window. It exposes a key-protected `POST /dsh-idle-watchdog/api/prompt` endpoint, enabling scripts and webhooks to deliver messages to sessions by id, prefix, or title. It supports optional strategies such as quiet hours, backoff, and goal/error awareness. Configuration is stored in settings or profile patches for DSH 0.1.5 and 0.2.x. Install with `dsh plugin --profile web add dsh-idle-watchdog`, then restart `dsh web` and enable it under Settings → Idle Watchdog.
Read Moredsh-fs-reobserve: Agent refreshes file observations after shell execution to avoid FS_STALE_VERSION
DSH's file observation strategy only learns file versions from instrumentation reads and writes. Consequently, modifications made via bash/shell do not trigger `fs/observed` reports, leading to false `FS_STALE_VERSION` alerts on the agent's subsequent edit attempts. `dsh-fs-reobserve` resolves this self-inflicted staleness by re-statting previously observed paths and emitting new `fs/observed` events immediately after the current bash/shell session concludes. This mechanism eliminates stale states resulting from self-operations while preserving those caused by third-party writes and maintaining `FS_NOT_OBSERVED` behavior. This article covers the installation command, `cordis.patch.yml` configuration options, and troubleshooting entry points.
Read MoreRX-GEO Basics-B9-Brand Safety & Reputation: Negative Propagation Monitoring and Crisis Response in the AI Search Environment
Released by the Baitong AI Service Platform, this professional Skill is located at `@baitongai/rx-geo-ji-chu-b9-pin-pai-an-quan-sheng-yu` (Version 1.0.0) and requires an API Key. It extends brand safety monitoring to AI search answer pages: by inputting monitoring data and insight reports on the proportion of negative sentiment, it outputs reputation emergency plans. The service covers five core capabilities: negative propagation monitoring, identification of competitor disparagement, three-tier crisis response (minor, medium, and major), official statements and crisis PR SOPs, and positive digital PR hedging. It forms a sister sentinel pair with B8, representing "misstatement vs. malicious sentiment." The official page does not provide an installation command; it must be obtained from the official address.
Read More