Claude ignores exact user-specified names and shortens/interprets them instead

Problem

When a user specifies an exact name — for a skill, command, file, or any artifact — Claude defaults to a shorter or interpreted version instead of using the user's exact words. This happens even when the user repeats the name multiple times.

Example from real session

User said "sync me up" as the skill name — stated clearly, 4 separate times across the conversation. Claude named the skill sync anyway on the first build, then had to be corrected repeatedly before using the correct name sync-me-up.

Why this matters

• Users should never have to repeat an explicit name more than once
• Claude should treat user-provided names as exact strings, not as suggestions to be interpreted or shortened
• This is especially harmful for ADHD users who already have to repeat themselves more than neurotypical users and should not have that burden compounded by an AI that ignores explicit instructions

Expected behavior

When a user says "call it X" or uses a name phrase, Claude uses that exact phrase — no shortening, no reinterpreting, no defaulting to a "cleaner" version — unless the user explicitly asks for a suggestion.

Submitted on behalf of

A Claude Code user (Oak Park Construction).