Agentic Engineering Patterns

Overview

Agentic Engineering Patterns is a reference concept for the practical design, implementation, and operation of agentic systems—software workflows where AI agents can plan, act, use tools, and iterate toward goals. In the newsletter, it is framed as a hub for concrete patterns that help teams move from vague experimentation to repeatable engineering practice, especially for coding-agent workflows such as red/green TDD.

For AI Product Managers, this matters because agentic products are rarely successful through model capability alone. They depend on interaction design, tool access, feedback loops, safety boundaries, and operational patterns that make agent behavior reliable enough for real use. A pattern library like this helps PMs translate emerging agent capabilities into product requirements, evaluation criteria, and team workflows that are easier to scale.

Key Developments

• 2026-02-23 — Mentioned as a guide collecting patterns for building and operating agentic systems. The newsletter described it as a hub for specific implementations, including red/green TDD for coding agents.
• 2026-02-27 — Referenced through advice to “hoard things you know how to do,” emphasizing that retaining knowledge of techniques, capabilities, and possible workflows improves how developers collaborate with coding agents.

Relevance to AI PMs

• Turn agent behavior into product specs: Pattern-based thinking helps PMs define how an agent should plan, use tools, ask for clarification, recover from failure, and signal progress—making requirements more testable than generic “AI assistant” goals.
• Improve reliability in coding and workflow automation: Patterns such as red/green TDD provide practical scaffolding for agent-assisted execution. PMs can use these patterns to shape evaluation benchmarks, rollout criteria, and human-in-the-loop checkpoints.
• Guide platform and tooling decisions: As teams consider browser tools, APIs, or structured tool-calling systems, pattern libraries provide a framework for deciding what capabilities an agent actually needs and how those capabilities should be exposed safely.

Related

• coding-agents — Closely connected because many of the referenced patterns are most concrete in software-development workflows, where agents write, test, and revise code.
• simon-willison — Associated with the guide and the surrounding commentary that frames agentic engineering as a practical discipline rather than a purely conceptual one.
• redgreen-tdd — A specific example of an agentic workflow pattern, illustrating how traditional engineering practices can be adapted for coding agents.
• webmcp — Related as an enabling mechanism for structured tool access, which is often necessary to make agentic patterns operational in real systems.