智能体
Agent 规划、多步流程与编排
refactoring-agent-instructions
bnadlerjr
Refactors bloated agent instruction files (CLAUDE.md, AGENTS.md, COPILOT.md, .cursorrules, etc.) to follow progressive disclosure — minimal root file linking to categorized files in guidelines/. Use when agent instructions are too long, hard to maintain, contain contradictions, or need reorganization.
copilot-vscode-docs
arisng
Comprehensive research and grounding of answers using official VS Code and GitHub Copilot documentation. Use before conducting web searches for VS Code features, Copilot capabilities, configuration, debugging, extensions, remote development, or any official Microsoft documentation queries.
arcblock-context
ArcBlock
Load ArcBlock company context (products, technical architecture, strategy) on demand. Use /arcblock-context to see available topics, or /arcblock-context <topic> to load specific context.
langgraph-multiagent
BjornMelin
Architect-level development, audit, and migration of multi-agent systems using LangGraph (v1+) and LangChain (v1+). Use when building or refactoring supervisor/subagent architectures, orchestrator-worker workflows, routing/hand-offs, agentic RAG, memory (short + long-term), state + context engineering, guardrails + human-in-the-loop, MCP tool integration, observability (LangSmith/OpenTelemetry), deployment, and performance/cost optimization — or when migrating off deprecated patterns like langgraph.prebuilt.create_react_agent and libraries like langgraph-supervisor(-py), LlamaIndex agents, CrewAI, Agno, or OpenAI Agents.
agent-browser
fellipeutaka
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction.
codex-sdk
BjornMelin
Architect-level guidance, workflows, and scripts for building agentic coding systems with OpenAI Codex. Use for Codex SDK (@openai/codex-sdk) threads + streaming JSONL events; Codex CLI automation (codex exec, output schema, JSONL, resume); MCP server usage (codex mcp-server) and dynamic tool integration; multi-agent orchestration with OpenAI Agents SDK (handoffs, gating, tracing); durable state, caching, and memory using SQLite; safe-by-default sandbox/approval patterns and execpolicy rules.
hookdeck
hookdeck
"Routes to the correct Hookdeck product skill. Covers Event Gateway (inbound: receive webhooks, webhook endpoints, local dev, event queue) and Outpost (outbound: send webhooks to customers). Use when working with any Hookdeck product and unsure which skill to use."
literature
mseok
"Academic literature discovery, synthesis, and bibliography management. Find papers, verify citations, create .bib files, download PDFs, and synthesize literature narratives. Includes OpenAlex API integration for structured scholarly queries."
copilot-sdk-dotnet
arisng
Build applications with GitHub Copilot CLI SDKs for .NET. Use for direct CopilotClient integration or Microsoft Agent Framework. Covers sessions, streaming, tools, MCP, permissions, and multi-agent workflows.
coding-workflow
bnadlerjr
Use when user asks to build a feature, implement something new, or make significant code changes. Recognizes requests like "build", "implement", "create a new feature", "add functionality", "develop", "I need to build X", "let's implement", "new feature request", "make these changes". Orchestrates a four-stage workflow (Research → Brainstorm → Plan → Implement) using the appropriate thought pattern skill at each stage.
divide-and-conquer
build000r
Decompose complex tasks into independent, parallel sub-agents with no overlapping concerns, race conditions, or inter-agent dependencies. Autonomous execution: plans the split, launches immediately, then runs a Codex review agent to verify, fix, and commit. Use BEFORE spawning Task agents when the work involves multiple files, domains, or concerns that could be parallelized. Triggers on: planning multi-agent work, "split this into agents", "parallelize this", "divide and conquer", or when about to launch 2+ Task agents for a complex task. Also use when a task feels too large for a single agent but the right decomposition isn't obvious.
ai-sdk-agents
BjornMelin
Expert guidance for building AI agents with ToolLoopAgent (AI SDK v6+). Use when creating agents, configuring stopWhen/prepareStep, callOptionsSchema/prepareCall, dynamic tool selection, tool loops, or agent workflows (sequential, routing, evaluator-optimizer, orchestrator-worker). Triggers: ToolLoopAgent, agent loop, stopWhen, stepCountIs, prepareStep, callOptionsSchema, prepareCall, hasToolCall, InferAgentUIMessage, agent workflows.
skill-creator
negz
Create new skills, modify and improve existing skills. Use when users want to create a skill from scratch, turn a workflow into a skill, update or optimize an existing skill, improve a skill's description for better triggering accuracy, or when they mention SKILL.md files.
agent-browser
markus1189
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.
research-ideation
mseok
"Generate structured research questions, testable hypotheses, and empirical strategies from a topic or dataset. Produces 3–5 ranked questions with identification strategies."
python
ahgraber
Python skill router. Use when planning, implementing, or reviewing Python changes and you need to select focused skills for workflow, design, typing/contracts, reliability, testing, data/state, concurrency, integrations, runtime operations, or notebook async behavior.
creating-agents
bnadlerjr
Expert guidance for creating, auditing, and improving Claude Code agent definition files (.md). Use when working with ~/.claude/agents/ files, authoring new sub-agents, improving existing agents, or understanding agent configuration and system prompt best practices.
commit-message
ahgraber
"Use when drafting a Conventional Commit message from currently staged Git changes. Triggers: 'write commit message', 'draft commit from staged changes', 'generate conventional commit', 'message for git commit -F -'. Not for creating commits, amending history, or full code review."
git-atomic-commit
arisng
'Analyze git changes, group into atomic commits, generate conventional commit messages with proper type(scope) format. Use when committing changes, grouping staged/unstaged files, or generating commit messages. Enforces universal commit types + repo-specific scopes from .github/git-scope-constitution.md.'
devils-advocate
mseok
"Challenge research assumptions and identify weaknesses in arguments. Stress-test papers before submission or revision."
thinking-patterns
bnadlerjr
Provides structured reasoning patterns that produce visible, auditable output. Use when the user asks to "think through", "reason about", "work through", "show your thinking", or wants step-by-step reasoning. Also invoked with /thinking or /thinking <pattern>.
skill-manager
ewanyuan
Provides unified configuration and log storage service for other skills, supporting data sharing and collaboration between skills
cox
ewanyuan
Cox 编程 - 为您的 AI 编程体验保驾护航。帮助开发团队掌握项目进展、识别开发风险、了解系统健康状态。提供项目进度跟踪、迭代管理（MVP驱动）、任务状态管理、开发假设记录、应用模块监控、测试埋点和异常分析等功能。支持静态网页和交互网页两种方案，适合不同环境和团队规模。网页按迭代分组展示，清晰呈现每个迭代的进度和任务。
cox
ewanyuan
Cox Programming - Safeguarding your AI programming experience. Helps development teams grasp project progress, identify development risks, and understand system health status. Provides project progress tracking, iteration management (MVP-driven), task status management, development assumption recording, application module monitoring, test tracing points, and anomaly analysis functions. Supports both static web and interactive web solutions, suitable for different environments and team sizes. Web pages are grouped by iteration, clearly presenting the progress and tasks of each iteration.