# Agents and harnesses

> Free resources, with credit to the original creators.

## Resources
- [AGENTS.md](/e/agentsmd-agents-md/index.md): Open format for a repo instruction file that coding agents read. (beginner; by AGENTS.md (multi-vendor open format))
- [Building Effective AI Agents](/e/anthropic-building-effective-agents/index.md): Workflows vs agents and the core patterns; argues to start simple. Readable without code. (beginner; by Anthropic)
- [AI Agents Course](/e/hf-ai-agents-course/index.md): Free course from 'what is an agent' to building one. (beginner; by Hugging Face)
- [A practical guide to building agents (PDF)](/e/openai-practical-guide-to-building-agents/index.md): Plain-language foundations: when to build an agent, orchestration, guardrails. (beginner; by OpenAI)
- [Equipping agents for the real world with Agent Skills](/e/anthropic-agent-skills/index.md): Packaging instructions and scripts as loadable skills. (intermediate; by Anthropic)
- [Best practices for Claude Code](/e/anthropic-claude-code-best-practices/index.md): Concrete coding-agent workflow habits: plan, verify, manage context. (intermediate; by Anthropic)
- [Effective context engineering for AI agents](/e/anthropic-effective-context-engineering/index.md): Treats context as a finite budget: compaction, notes, sub-agents. (intermediate; by Anthropic)
- [12-Factor Agents](/e/humanlayer-12-factor-agents/index.md): Principles for agents that hold up in production. (intermediate; by HumanLayer (Dex Horthy))
- [Agents](/e/huyenchip-agents/index.md): Long-form essay on tools, planning, failure modes and evaluating agents. (intermediate; by Chip Huyen)
- [LLM Powered Autonomous Agents](/e/lilianweng-llm-powered-autonomous-agents/index.md): Classic overview of planning, memory and tool use in agents. (intermediate; by Lilian Weng)
- [Model Context Protocol docs](/e/mcp-model-context-protocol-docs/index.md): The spec and docs for connecting tools and data to agents. (intermediate; by MCP project)
- [OpenAI Agents SDK](/e/openai-agents-sdk/index.md): Lightweight framework: handoffs, guardrails, tracing. (intermediate; by OpenAI)
- [Context Engineering for Agents](/e/rlancemartin-context-engineering-for-agents/index.md): Write/select/compress/isolate taxonomy of context strategies. (intermediate; by Lance Martin)
- [Effective harnesses for long-running agents](/e/anthropic-effective-harnesses-long-running-agents/index.md): Harness design for agents that work across many context windows. (advanced; by Anthropic)
- [How we built our multi-agent research system](/e/anthropic-multi-agent-research-system/index.md): Real production lessons on orchestrator/sub-agent design and its costs. (advanced; by Anthropic)
- [Writing effective tools for AI agents](/e/anthropic-writing-tools-for-agents/index.md): How to design, evaluate and improve the tools an agent calls. (advanced; by Anthropic)
- [Context Rot: How Increasing Input Tokens Impacts LLM Performance](/e/chroma-context-rot/index.md): Measures how output quality drops as input length grows. (advanced; by Chroma)
- [Don't Build Multi-Agents](/e/cognition-dont-build-multi-agents/index.md): The counter-argument: context sharing beats splitting work across agents. (advanced; by Cognition)
- [Context Engineering for AI Agents: Lessons from Building Manus](/e/manus-context-engineering-lessons/index.md): KV-cache-aware design, file system as memory, keep errors in context. (advanced; by Manus)
- [MCP reference servers](/e/mcp-reference-servers/index.md): Reference implementations to learn MCP server design from. (advanced; by MCP project)
- [OpenHands](/e/openhands-openhands/index.md): Open-source coding-agent platform worth reading as a reference harness. (advanced; by OpenHands (All Hands AI))
- [The lethal trifecta for AI agents](/e/willison-lethal-trifecta/index.md): A defensive framework for recognizing risky combinations of agent capabilities and untrusted inputs. (advanced; by Simon Willison)
- [Design Patterns for Securing LLM Agents against Prompt Injections](/e/willison-prompt-injection-design-patterns/index.md): Summary of a paper's defensive agent architectures. (advanced; by Simon Willison)
- [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection](/e/greshake-indirect-prompt-injection/index.md): Foundational public research on indirect prompt injection. Linked only; no payloads reproduced. (expert; by Kai Greshake et al.)
- [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](/e/yang-swe-agent/index.md): Research on how the agent-computer interface influences software-engineering performance. (expert; by John Yang et al.)
- [ReAct: Synergizing Reasoning and Acting in Language Models](/e/yao-react/index.md): Research on combining reasoning with environment actions, a useful reference for agent-loop design. (expert; by Shunyu Yao et al.)