Formal Architecture Descriptors as Navigation Primitives for AI Coding Agents

TL;DR

Providing AI coding agents with formal architecture descriptors significantly reduces navigation overhead, improves accuracy, and enhances consistency in code exploration tasks.

Contribution

The paper introduces intent.lisp, an S-expression architecture descriptor, and demonstrates its effectiveness through experiments and a field study, advancing AI code navigation methods.

Findings

Architecture context reduces navigation steps by 33-44%.

Automatically generated descriptors achieve 100% accuracy.

52% reduction in behavioral variance in field sessions.

Abstract

AI coding agents spend a substantial fraction of their tool calls on undirected codebase exploration. We investigate whether providing agents with formal architecture descriptors can reduce this navigational overhead. We present three complementary studies. First, a controlled experiment (24 code localization tasks x 4 conditions, Claude Sonnet 4.6, temperature=0) demonstrates that architecture context reduces navigation steps by 33-44% (Wilcoxon p=0.009, Cohen's d=0.92), with no significant format difference detected across S-expression, JSON, YAML, and Markdown. Second, an artifact-vs-process experiment (15 tasks x 3 conditions) demonstrates that an automatically generated descriptor achieves 100% accuracy versus 80% blind (p=0.002, d=1.04), proving direct navigational value independent of developer self-clarification. Third, an observational field study across 7,012 Claude Code sessions shows 52% reduction in agent behavioral variance. A writer-side experiment (96 generation runs, 96 error injections) reveals critical failure mode differences: JSON fails atomically, YAML silently corrupts 50% of errors, S-expressions detect all structural completeness errors. We propose intent.lisp, an S-expression architecture descriptor, and open-source the Forge toolkit.