Agent Design Patterns: 510 curated resources
Curated material on agent design patterns within agents and harnesses.
510 resources · page 16 of 16
Metaflow documentation
Creator not supplied; see original source.
Metaflow documentation: open-source framework from Netflix for building and running data science and ML workflows.
Nanbeige 4.2-3B official model card
Creator not supplied; see original source.
Hugging Face model card for Nanbeige4.2-3B with architecture, usage and evaluation details.
NAVER D2
Creator not supplied; see original source.
NAVER D2 Hello World: Korean engineering blog from NAVER developers.
Notion Help: Guides
Creator not supplied; see original source.
Notion Help Center guides to workflows, integrations and features.
OpenAI API Skills guide
Creator not supplied; see original source.
OpenAI API skills guide: creating reusable skills for Responses API shell tools and Agents API sandboxes.
PennyLane documentation
Creator not supplied; see original source.
PennyLane documentation: Python library for differentiable quantum computing and hybrid quantum-classical ML.
PyGOD
Creator not supplied; see original source.
PyGOD documentation: Python library for outlier and anomaly detection on graphs.
ReAct
Creator not supplied; see original source.
arXiv paper (2022) "ReAct: Synergizing Reasoning and Acting in Language Models", introducing ReAct, interleaving reasoning traces and tool actions in language models.
Refactoring Hermes with 1,393 agents
Creator not supplied; see original source.
Nous Research case study of Hermes Agent refactoring its own codebase with 1,393 subagents, with cost and results.
Roo Code docs: Context poisoning
Creator not supplied; see original source.
Roo Code documentation on context poisoning: recognising and recovering from corrupted context in long agent sessions.
Scaling Managed Agents: decoupling brain from hands
Creator not supplied; see original source.
Anthropic engineering post on the architecture of its hosted Managed Agents harness, separating model reasoning from execution.
SHAP documentation
Creator not supplied; see original source.
SHAP documentation: game-theoretic method and library for explaining ML model predictions.
Building production-ready agentic systems: Lessons from Shopify Sidekick
Creator not supplied; see original source.
Shopify Engineering post on lessons from building the Sidekick agent: architecture, LLM judges calibrated to human labels and RL pitfalls.
Durable AI (Temporal Documentation)
Creator not supplied; see original source.
Temporal documentation on durable execution for AI applications and agents that resume after crashes or timeouts.
Texterra
Creator not supplied; see original source.
Texterra from ISP RAS: Russian text-analysis system for information extraction built on open structured knowledge.
Training language models to follow instructions with human feedback
Creator not supplied; see original source.
arXiv paper (2022) "Training language models to follow instructions with human feedback", describing InstructGPT and the RLHF post-training pipeline.
TRL documentation
Creator not supplied; see original source.
Hugging Face TRL documentation: trainers for SFT, DPO, GRPO and other post-training methods.
Universal Commerce Protocol (UCP)
Creator not supplied; see original source.
Universal Commerce Protocol site: open standard for agents and platforms covering catalogue, checkout and orders.
AI SDK docs: Loop control
Creator not supplied; see original source.
Vercel AI SDK documentation on loop control: stop conditions and step control for agent loops.
How we made v0 an effective coding agent (Vercel)
Creator not supplied; see original source.
Vercel engineering post on v0's coding-agent pipeline: dynamic system prompts, streaming fixes and autofixers.
ReAct: Synergizing Reasoning and Acting in Language Models
Research on combining reasoning with environment actions, a useful reference for agent-loop design.
Barry Sookman: Moffatt v Air Canada
Creator not supplied; see original source.
Barry Sookman's analysis of Moffatt v. Air Canada, the case holding a company liable for its chatbot's false statements.
Berkeley Advanced LLM Agents (SP25)
Creator not supplied; see original source.
Berkeley CS294/194-280 Advanced Large Language Model Agents (Spring 2025): lectures on reasoning, coding agents and agent safety.
Brynjolfsson, Li, Raymond: Generative AI at Work (NBER w31161)
Creator not supplied; see original source.
NBER working paper w31161 Generative AI at Work, measuring productivity effects of an AI assistant on customer support agents.
EASA AI Concept Paper Issue 2
Creator not supplied; see original source.
EASA AI Concept Paper Issue 2: guidance on assuring Level 1 and 2 machine-learning applications in aviation.
GEO: Generative Engine Optimization
Creator not supplied; see original source.
arXiv HTML version of the GEO paper (2023) defining generative engine optimization and its visibility metrics.
Model Context Protocol Specification 2025-06-18
Creator not supplied; see original source.
MCP specification version 2025-06-18, covering authorisation, elicitation and structured tool output.
OpenTelemetry generative AI semantic conventions
Creator not supplied; see original source.
OpenTelemetry page on generative AI semantic conventions for LLM and agent telemetry, now moved to a dedicated repository.
SpikingJelly
Creator not supplied; see original source.
SpikingJelly documentation (in Chinese): PyTorch framework for deep learning with spiking neural networks.
Tülu 3
Creator not supplied; see original source.
arXiv paper (2024) describing Tülu 3, a fully open post-training recipe that introduced RLVR.