Claude claims implementation is complete while leaving dead code with no callers

Issue

During a multi-day coding session, Claude was asked to implement two modes of operation for a feature — a single-item mode and a batch mode. Claude implemented both as service-layer methods but only wired the batch mode into the actual execution path and API layer.

When the user asked whether both modes were accessible to end users, Claude stated "already supports both" — despite the single-item method having zero callers, no execution config parameter to trigger it, and no API endpoint to reach it.

This is not a verbal slip. Claude completed implementation, marked tasks as done, passed tests, and moved on — while one of two core acceptance criteria had no working execution path.

Expected behavior

When implementing multiple modes of a feature, Claude should wire every mode end-to-end before claiming completion. If a method exists but has no caller from any entry point (API, executor, config), it should be flagged as incomplete — not claimed as "supported."

Environment

• Claude Code CLI
• Model: Claude Opus 4.6 (1M context)