skills.rest
tankpkg
tankpkgOFFICIAL
Explore all skills by this author →
GitHub

@tank/ai-agent-patterns

Design production AI agents with tool use, memory, safety, and observability.

1|1|Updated Feb 20, 2026
Search Tags:#ai agents#tool calling#multi-agent#observability#memory systems#guardrails#agent architecture
One-click install
npx skills add https://github.com/tankpkg/packages --skill tank-ai-agent-patterns
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: @tank/ai-agent-patterns
Source: https://github.com/tankpkg/packages/tree/main/skills/ai-agent-patterns
Command: npx skills add https://github.com/tankpkg/packages --skill tank-ai-agent-patterns

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill helps you design, evaluate, and ship AI agents that can use tools reliably, stay within budget, and behave safely in production. It reduces the trial-and-error of choosing an agent architecture and gives you a practical path from prototype to durable system.

Core Features & Use Cases

  • Architecture selection: Choose between ReAct, Plan-and-Execute, Reflexion, LATS, and multi-agent orchestration based on task shape, cost, and reliability needs.
  • Tool calling and orchestration: Define robust tools, handle structured output, parallel calls, retries, and dependency-aware execution.
  • Memory, oversight, and safety: Add sliding window, semantic, episodic, and working memory; insert approval gates and interrupts; apply guardrails, tracing, evaluation, and token budgets.
  • Use Case: If you are building a customer-support or research agent, this Skill helps you decide the right framework, structure the workflow, prevent runaway costs, and instrument the system for review and regression testing.

Quick Start

Ask for the best production agent architecture for your task and have it map the workflow, tools, memory, guardrails, and evaluation plan.

Frequently Asked Questions about @tank/ai-agent-patterns

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I choose the best AI agent architecture for my task?▼

Choose an AI agent architecture by matching task shape, cost, and reliability needs against patterns like ReAct, Plan-and-Execute, Reflexion, LATS, or multi-agent orchestration.

How do I build reliable tool calling and orchestration for AI agents?▼

Build reliable tool calling by defining robust tools, handling structured output, parallel calls, retries, and dependency-aware execution to manage complex agent workflows safely.

Does this approach work with LangGraph, CrewAI, Mastra, and OpenAI Agents SDK?▼

Yes, this architecture selection, multi-agent orchestration, and tool use approach applies directly to LangGraph, CrewAI, Mastra, OpenAI Agents SDK, and Anthropic tool use workflows.

How to prevent runaway costs and add guardrails to production AI agents?▼

Prevent runaway costs and add guardrails by applying token budgets, human approval gates, interrupts, tracing, and evaluation to keep agent behavior robust under real-world conditions.

What memory systems should I use for a durable AI agent system?▼

Use sliding window, semantic, episodic, and working memory systems to maintain context and state, ensuring your production AI agent behaves reliably across long-running sessions.

Why do my multi-agent workflows fail under real-world conditions?▼

Multi-agent workflows fail without structured tool schemas, routing, caching, checkpointing, and evaluation, which are required to keep agent behavior robust and prevent errors in production.

Related Skills

Explore More Software Engineering Skills→
ricardorochericardoroche
2

agent-orchestration-patterns

Standardize Pydantic tool schemas and agent state for multi-agent orchestration.

Community
Intermediate
MattMaggMattMagg
3

ADK Production

Deploy and operate ADK agents on Vertex AI Agent Engine, Cloud Run, and GKE.

Community
Advanced
consuelohqconsuelohq
7

task-state-manager

Persist task state in Linear and trace execution with LangSmith.

Official
Advanced
faionfaionfaionfaion
4

faion-ai-agents

Guide AI agent development with LangChain and LlamaIndex architectures.

Community
Advanced