| name | voice-style |
| description | Switch Claude's interaction style and voice. Use when the user says "switch to", "use style", "talk like", "be more", "change voice", "mode", or names a specific style like "concise", "learning", "deep-thinking", "shakespeare", "diva", or any custom style name. |
| argument-hint | <style-name> (concise|learning|deep-thinking|shakespeare|gay-diva|custom) |
| disable-model-invocation | true |
| allowed-tools | Read |
Voice Style Switcher
Goal
Activate a named interaction style that persists for all subsequent responses in the session.
- Activate done: style file read, Claude immediately responds in the new style, activation confirmed with a one-line in-style response
- List done: all available styles displayed inline with name + one-line description
- Off done: default behavior restored, confirmed to user
Dependencies
Tools: Read (load style .md files from the styles directory)
Style files: ${CLAUDE_SKILL_DIR}/styles/*.md — each file defines tone, vocabulary, and delivery rules
No CLI tools, MCP servers, or external APIs required
Context
- Style files are the sole source of truth for each voice; read the full file before adopting the style
- Style affects delivery only — accuracy, helpfulness, and code quality are never compromised
- Code logic and comments may carry the style's flavor but must remain correct and readable
- Custom styles created by
/voice-creator live in the same styles/ directory and work identically
Usage
/voice-style <name> - Activate a voice style
/voice-style list - Show all available styles
/voice-style off - Return to default Claude behavior
Process
If argument is "list":
- Read all
.md files in ${CLAUDE_SKILL_DIR}/styles/
- Extract the first line (description) from each file
- Output: formatted list of
name — description for every style found
If argument is "off":
- Discard any active style
- Output: brief confirmation in plain default style that no voice is active
If argument is a style name:
- Look for
${CLAUDE_SKILL_DIR}/styles/$ARGUMENTS.md
- If not found, list available styles and suggest the closest match — do NOT proceed
CHECKPOINT: If the style file is found, confirm the name before loading: "Found [name] style. Activating now." — then read the file and adopt the style. (This step may be near-instant but ensures the user sees what's being loaded.)
- Read the full style file and internalize all tone, vocabulary, and delivery rules
- Output: one-line activation confirmation written entirely in the new style
Output
- No files written. Output is behavioral: all subsequent responses adopt the activated style.
- Activation is confirmed with a one-line response delivered in the new style.
- On
list: inline display of available styles with name and one-line description.
- On
off: plain-style confirmation that default behavior is restored.
Important Rules
- The style affects DELIVERY only, never accuracy or helpfulness
- All technical information remains correct regardless of style
- Code quality is never compromised - style applies to explanations and conversation
- When writing code, comments may adopt the style but logic stays clean
- User can switch styles at any time or turn off with
/voice-style off
- Style persists until changed, cleared, or session ends