Analyze one or multiple images with Gemini Flash (vision) via the generateContent REST API. Uses Bun + TypeScript to base64-encode images and print model output. Use for captioning, OCR-like extraction, UI/screenshot analysis, and multi-image comparison.
Generate and execute SQL queries against the civitai_records PostgreSQL database to analyze video performance on Civitai. Use when users ask about: video engagement metrics (likes, hearts, comments), content performance analysis, tag/theme analysis, quality scores, weekly reports, comparing videos, content recommendations, trend analysis, or any Civitai data queries. Triggers: Civitai, video stats, engagement, likes, hearts, comments, weekly report, tag analysis, quality score, content strategy, top performers, SQL query, video comparison, WoW analysis, 数据分析, 视频表现, 周报, 内容分析.
Build, create, debug, review, implement, and optimize agentic AI applications using the OpenAI Agents SDK for TypeScript. Use when creating new agents, defining tools, implementing handoffs between agents, adding guardrails, debugging agent behavior, reviewing agent code, or orchestrating multi-agent systems with the @openai/agents package.
Design effective tools for AI agents - tool descriptions, consolidation principles, architectural reduction, MCP naming conventions, and error handling
Automatic design system context injection for UI consistency
Enable voice conversations with Claude Code - speak commands and hear responses
Embassy HAL for Raspberry Pi RP2040 and RP2350 microcontrollers. Use when working with Pico/Pico 2 peripherals, GPIO, SPI, I2C, UART, PIO, or flashing with picotool.
Configure and optimize AI coding agents (OpenCode/Claude). Use when setting up or improving agent behavior.
Generate adversarial inputs, edge cases, and boundary test payloads for stress-testing LLM robustness
プロジェクトの CLAUDE.md を更新する。「CLAUDE.md を更新して」「ルールを追加」「プロジェクトルールを変更」「Claude のルール」「AI ドキュメント更新」「CLAUDE.md に追記」「ルールを編集」などで起動。プロジェクト固有のルールやガイドラインを管理。
Use when subagents need to delegate work but can't use Task tool, or when needing to test skills in isolated context - spawns Claude instances via CLI backgrounding with JSON responses
RAG pipeline, embeddings, LLM interactions, and flow orchestration.
Use when building retrieval-augmented generation systems. Covers chunking strategies, embedding models, vector databases, retrieval patterns, and hybrid search. Apply when adding knowledge bases, document Q&A, or semantic search to applications.
Guide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).
Use when testing Claude Code plugins programmatically, building SDK applications, encountering plugin loading failures (plugins array empty), getting AttributeError on message.type, or needing verified Claude Agents SDK Python patterns - provides complete reference for correct message handling (isinstance not .type), plugin loading (setting_sources required), content block iteration, and all working patterns from official documentation
AI产品经理,整套"0基础 AI 协同开发系统"的源头角色与项目总导演。将自然语言需求转换为可执行的产品规划与标准化任务卡,协调 Backend/Frontend/SCF/QA/Reviewer/Deploy/Billing Guard 多部门有序并行交付。遵循易用性优先、契约优先、门禁治理、MVP-first 的工程基线。适用于收到自然语言需求时使用。
Design and evaluate context compression strategies for long-running agent sessions. Use when agents exhaust memory, need to summarize conversation history, or when optimizing tokens-per-task rather than tokens-per-request.
Initialize project.faf files when starting new projects, when user asks to set up AI context, create project DNA, or needs persistent context. Automatically detects project type (React, TypeScript, Next.js, Svelte, etc.), generates appropriate IANA-registered format (application/vnd.faf+yaml), and creates project.faf in repository root. Use when user mentions "context", "project setup", "AI-readiness", or "FAF".
U-llm-sdk and claude-only-sdk patterns. Use when working on projects with LLM service, designing LLM integrations, or implementing AI-powered features.
NVIDIA Isaac Sim and Isaac ROS integration patterns for Physical AI applications.
Use when facing 3+ independent failures that can be investigated without shared state or dependencies - dispatches multiple Claude agents to investigate and fix independent problems concurrently
