发现 Skills

A Socratic coaching agent that helps users think through problems, decisions, and ideas by asking probing questions instead of giving direct answers. Use this skill whenever the user says "coach me", "help me think through", "Socratic mode", "challenge my thinking", "play devil's advocate", or when they're wrestling with a decision, strategy, product idea, or complex problem and would benefit from guided questioning rather than direct advice. Also trigger when the user mentions "Socrates" or "probing questions" in any context.

First-run orientation for a new BenchAGI teammate or pilot customer. Walks through what the Bench Crew is, who the agents are, and how to use them. Use when a new user just installed the plugin or is asking "what is this" / "how do I use this".

Choose and implement clustering algorithms for grouping speaker embeddings after VAD and embedding extraction. Compare Hierarchical clustering (auto-tunes speaker count), KMeans (fast, requires known count), and Agglomerative clustering (fixed clusters). Use Hierarchical clustering when speaker count is unknown, KMeans when count is known, and always normalize embeddings before clustering.

Initialize a complete JavaScript/TypeScript project with package.json, testing framework, bundler configuration, and documentation. Use when starting Node.js apps, React projects, Express APIs, or npm packages.

Design and implement durable, fault-tolerant AI agent pipelines using Temporal workflow orchestration. Covers when to use Temporal vs simple async code, TypeScript SDK setup, AI agent activity design, parallel agent fan-out, crash recovery, and Docker Compose deployment. Use when building multi-agent AI systems that must survive failures, run in parallel, and resume after crashes.

Guided, section-by-section authoring of the master architecture document for the game. Reads all GDDs, the systems index, existing ADRs, and the engine reference library to produce a complete architecture blueprint before any code is written. Engine-version-aware: flags knowledge gaps and validates decisions against the pinned engine version.

Analyzes what is done and the user's query and offers advice on what to do next. Use this if the user asks “What should I do next?”, “What do I do now?”, “I'm stuck”, or “I don't know what to do.”

Complete launch readiness validation covering every department: code, content, store, marketing, community, infrastructure, legal, and go/no-go sign-offs.

Generate player-facing patch notes from git history, sprint data, and internal changelogs. Translates developer language into clear, engaging player communication.

Advanced swarm orchestration patterns for research, development, testing, and complex distributed workflows

Use Haze features for blur effects in Compose Multiplatform. Covers hazeSource, hazeEffect, blurEffect, styling, progressive blurs, masking, overlapping effects, and common UI patterns.

Use when reviewing accessibility of websites, apps, or facilities, conducting WCAG audits, preparing for ADA Title II/III deadlines, or checking digital/physical accessibility requirements.

Log all chat messages to a SQLite database for searchable history and audit. Use when: (1) Building chat history, (2) Auditing conversations, (3) Searching past messages, or (4) User asks to log chats.

AI operational modes (brainstorm, implement, debug, review, teach, ship). Adapts behavior based on task type for Trust-Grade execution.

Context window budget management and selective capability loading for LLM token optimization.

Extract reusable patterns from sessions and save them as knowledge

Internationalization and localization patterns for multi-language applications

Comprehensive verification running all quality gates

Attach a DuckDB database file for use with /duckdb-skills:query. Explores the schema (tables, columns, row counts) and writes a SQL state file so subsequent queries can restore this session automatically via duckdb -init.

Control and interact with a live browser session on any scraped page — click buttons, fill forms, navigate flows, and extract data using natural language prompts or code. Replaces the old firecrawl-browser command. Use when the user needs to interact with a webpage beyond simple scraping: logging into a site, submitting forms, clicking through pagination, handling infinite scroll, navigating multi-step checkout or wizard flows, or when a regular scrape failed because content is behind JavaScript interaction. Also useful for authenticated scraping via profiles. Triggers on "browser", "instruct", "click", "fill out the form", "log in to", "sign in", "submit", "paginated", "next page", "infinite scroll", "interact with the page", "navigate to", "open a session", or "scrape failed".

Use when building React 18+ applications in .jsx or .tsx files, Next.js App Router projects, or create-react-app setups. Creates components, implements custom hooks, debugs rendering issues, migrates class components to functional, and implements state management. Invoke for Server Components, Suspense boundaries, useActionState forms, performance optimization, or React 19 features.

Self-improvement framework for Claude Code prompts. Pass a profile name to evaluate variations of that target prompt via a 3-level cascade (stability via embeddings, decision-consistency via parsing, and pairwise quality via a judge bracket). The skill itself is the team lead — it dispatches all runner teammates directly at the top level (no nested teams), so every parallel agent is visible in the Claude Code Agent Teams tmux split.

花叔Design（Huashu-Design）——用HTML做高保真原型、交互Demo、幻灯片、动画、设计变体探索+设计方向顾问+专家评审的一体化设计能力。HTML是工具不是媒介，根据任务embody不同专家（UX设计师/动画师/幻灯片设计师/原型师），避免web design tropes。触发词：做原型、设计Demo、交互原型、HTML演示、动画Demo、设计变体、hi-fi设计、UI mockup、prototype、设计探索、做个HTML页面、做个可视化、app原型、iOS原型、移动应用mockup、导出MP4、导出GIF、60fps视频、设计风格、设计方向、设计哲学、配色方案、视觉风格、推荐风格、选个风格、做个好看的、评审、好不好看、review this design。**主干能力**：Junior Designer工作流（先给假设+reasoning+placeholder再迭代）、反AI slop清单、React+Babel最佳实践、Tweaks变体切换、Speaker Notes演示、Starter Components（幻灯片外壳/变体画布/动画引擎/设备边框）、App原型专属守则（默认从Wikimedia/Met/Unsplash取真图、每台iPhone包AppPhone状态管理器可交互、交付前跑Playwright点击测试）、Playwright验证、HTML动画→MP4/GIF视频导出（25fps基础 + 60fps插帧 + palette优化GIF + 6首场景化BGM + 自动fade）。**需求模糊时的Fallback**：设计方向顾问模式——从5流派×20种设计哲学（Pentagram信息建筑/Field.io运动诗学/Kenya Hara东方极简/Sagmeister实验先锋等）推荐3个差异化方向，展示24个预制showcase（8场景×3风格），并行生成3个视觉Demo让用户选。**交付后可选**：专家级5维度评审（哲学一致性/视觉层级/细节执行/功能性/创新性各打10分+修复清单）。

Run a full dealradar Phase A + Phase B cycle — generate targeted niche queries via the strategist sub-skill, scrape 10–15 keyword searches across LBC/Vinted/Wallapop, triage up to 200 listings, produce BUY/VERIFY/PASS verdicts with margin_net and velocity_days_est included in every payload, persist raw LLM verdicts to the dealradar API, and push Telegram top-deal alerts.