Day 3: The Architecture
Key question: How do we structure code for agents?
Today's ACIs
| ACI | Name | Takeaway |
|---|---|---|
| ACI-002 | CLAUDE.md as Living Architecture Document | CLAUDE.md is not a configuration file you write once. It evolves with the project. Edit frequency correlates with project health. |
| ACI-008 | Style Rules vs Architecture Rules | Style rules do not improve agent output. Architecture rules do. The filter: would violating this rule cause a bug or just a formatting inconsistency? |
| ACI-006 | The Great Simplification | CLAUDE.md files grow. At some point, they need to shrink. Distil to the rules that actually affect agent behaviour. |
| ACI-001 | The Highlander Rule | There can be only one. Agents create duplication because they solve problems locally. Never duplicate code paths for the same concern. |
| ACI-011 | Namespace Archaeology | Namespace design for agents is different from namespace design for humans. Agents navigate by reading, not by memory. |
| ACI-016 | Skills as Procedural Memory | Skills are markdown files that teach the agent how to perform specific tasks. Write it once, reference it forever. |