Lexical Anchors and Repository Graphs for Coding Agents

Abstract

This evidence synthesis examines anchored structural retrieval in coding agents that must locate relevant files and tests in unfamiliar repositories. It argues that the appropriate object of evaluation is the chain from query and index state through candidate retrieval, structural expansion, generation or extraction, confidence, and downstream use, not a model score in isolation. The review brings together the assigned studies with established work on uncertainty, robustness, provenance, and governance. Across these literatures, a common problem emerges: a fluent answer or entity label can rest on stale, temporally invalid, or systematically omitted evidence. The proposed framework separates evidence quality, model behavior, decision policy, and operational monitoring, then asks how each layer changes under distribution shift, adversarial pressure, or incomplete information. It recommends evaluation by slices and repeated trials, explicit reject and escalation policies, preservation of data and reasoning lineage, and prospective monitoring tied to defined actions. The result is a research agenda for systems that are efficient enough to use but also bounded enough to audit. No new experiment is claimed; the article develops a comparative conceptual model and identifies tests that would make future empirical claims more credible.