Agent building collection

AI Agent Development Skills for Claude Code and Antigravity

For developers building their own agents or LLM features who want the coding agent to follow proven patterns for tools, memory, evaluation and retrieval. To install one, download it and put the skill folder (the one with SKILL.md) in your agent’s skills directory: .claude/skills/ for Claude Code or .agents/skills/ for Antigravity. The install guide covers global folders and CLI options.

Updated November 2026

At a glance

SkillBest forAuthor
AI Agents ArchitectAgent architectureAntigravity Skills
AI Agent DevelopmentCrewAI and LangGraph buildsAntigravity Skills
Agent Tool BuilderTool and function designAntigravity Skills
Agent Memory SystemsDesigning agent memoryAntigravity Skills
Conversation MemoryChat memory with Mem0 or RedisAntigravity Skills
Agent Memory MCPPersistent memory over MCPAntigravity Skills
Context Window ManagementLong-context cost and qualityAntigravity Skills
Agent EvaluationTesting agent behaviourAntigravity Skills
Advanced EvaluationLLM-as-a-judge pipelinesAntigravity Skills
Claude APIBuilding on the Claude APIAntigravity Skills
AI EngineerProduction LLM apps and RAGAntigravity Skills
AI/ML DeveloperRAG with vector databasesAntigravity Skills

Which agent skill fits your job?

The 12 agent development skills

  1. Published by Antigravity Skills · Agent architecture

    Designs autonomous agents: ReAct loops, memory systems and multi-agent orchestration, aimed at reliable, controllable behaviour.

    Use it when: You are designing an agent and need to decide its loop, memory and orchestration up front.

  2. Published by Antigravity Skills · CrewAI and LangGraph builds

    Builds agents in four phases, from design to a single agent, multi-agent systems and orchestration, with CrewAI and LangGraph.

    Use it when: You are building a multi-agent system with CrewAI or LangGraph and want a phased plan.

  3. Published by Antigravity Skills · Tool and function design

    Designs the tools an agent calls so they use fewer tokens and fail loudly instead of silently.

    Use it when: Your agent picks the wrong tool, burns tokens or fails without saying why.

  4. Published by Antigravity Skills · Designing agent memory

    Covers short-term, long-term and cognitive memory architectures, with retrieval-first strategies against agent forgetting.

    Use it when: Your agent loses track of earlier facts and you need to choose a memory design.

  5. Published by Antigravity Skills · Chat memory with Mem0 or Redis

    Builds short-term, long-term and entity memory for agents using Mem0, LangChain and Redis.

    Use it when: You are implementing memory for a chat agent on Mem0, LangChain or Redis.

  6. Published by Antigravity Skills · Persistent memory over MCP

    A hybrid MCP memory server that gives agents persistent, searchable knowledge about architecture, patterns and decisions.

    Use it when: You want memory that any MCP-capable agent can read and write.

  7. Published by Antigravity Skills · Long-context cost and quality

    Manages LLM context in tiers with summarization and routing to avoid context rot and cut token costs.

    Use it when: Long sessions get slower, pricier and less accurate as the context fills up.

  8. Published by Antigravity Skills · Testing agent behaviour

    Evaluates agents with behavioural tests, capability assessment and reliability metrics.

    Use it when: You need evidence that an agent works before you ship it or change its prompt.

  9. Published by Antigravity Skills · LLM-as-a-judge pipelines

    Builds LLM evaluation pipelines with direct scoring and pairwise comparison, including bias mitigation.

    Use it when: You want automated quality scores for model outputs that you can trust.

  10. Published by Antigravity Skills · Building on the Claude API

    Integrates Claude into apps through the SDK, including adaptive thinking and streaming.

    Use it when: Your app calls Claude and you want the agent to use the SDK correctly.

  11. Published by Antigravity Skills · Production LLM apps and RAG

    Production LLM applications, RAG systems and AI agents, including vector search and agent orchestration.

    Use it when: You are taking an LLM prototype to production and need RAG and orchestration done properly.

  12. Published by Antigravity Skills · RAG with vector databases

    Builds production LLM apps with vector databases, RAG, prompt engineering and agentic workflows.

    Use it when: You need to pick a vector database and wire up retrieval for an LLM app.

How to install these skills

Put each skill folder (the one that contains SKILL.md) in your agent’s skills folder: .agents/skills/ in your project for Antigravity and Codex, or .claude/skills/ for Claude Code. The install guide covers global folders, CLI commands and a full worked example.

Read the install guide

Frequently asked questions

Which skill should I use to build an AI agent?
Start with AI Agents Architect to decide the loop, memory and orchestration, then add Agent Tool Builder for the tools and Agent Evaluation before you ship.
How is agent memory different from context management?
Memory skills decide what the agent stores and retrieves across turns or sessions. Context Window Management decides what goes into a single model call so it stays accurate and affordable.
Are these skills tied to one model provider?
Most are patterns that apply to any LLM. Claude API is specific to Anthropic’s SDK. Read each skill page for the libraries it uses.
How do I install these skills?
Download the skill and put its folder (the one that contains SKILL.md) in your agent’s skills directory: .claude/skills/ for Claude Code, or .agents/skills/ in your project for Antigravity and Codex. The install guide at /docs/install-skills lists global folders and CLI options.
What does a download cost?
Each packaged skill download on Antigravity Skills uses 1 credit. The skill pages and the install guide are free to read.
AI Agent Development Skills for Claude Code and Antigravity