42 Agent Architecture Patterns: From Skill Repos to Intent & Harness Engineering

From A Pattern Language of Agentic AI Skills Design https://intuitionmachine.gumroad.com/l/aiskills

The video presents an Alexandrian-style pattern language for Claude code development: named, interconnected solutions to recurring structural problems, emphasizing that prompting is insufficient and reliable agents require deterministic architecture through structural constraints and intent engineering. It outlines tensions such as context richness vs context window limits, autonomy vs safety, flexibility vs predictability, and speed vs correctness, then surveys patterns across multiple groups: structural foundations (flat skill catalog, skill directory anatomy, YAML front matter, trigger-style descriptions, knowledge workflow), context engineering (three-tier knowledge loading, write-to-file, source-attributed state, stable name references), harness engineering (explicit I/O contracts, pipeline composition, checkpointing, bounded correction loops, writer/evaluator separation, deterministic/creative split, agent-facing errors, graceful degradation), intent engineering (identity-first, iron laws, rationalization prevention, hard rules, constitution, philosophy injection), system/composition/testing/distribution/long-running patterns, culminating in guidance for scaling via disciplined structure, file-based data passing, and adversarial testing.

00:00 Pattern Language Overview
01:04 Structural Foundations
02:56 Context Engineering
05:26 Harness Engineering
07:58 Intent Engineering
10:53 System Engineering
12:55 Composition Patterns
14:01 Testing Patterns
15:23 Distribution Patterns
16:10 Long Running Agents
19:15 Wrap Up and Subscribe