Voice Mirror is a skill that maintains a living voice profile to make AI-generated text match your personal writing style. You provide writing samples such as emails, messages, or documents, and the skill extracts patterns including sentence rhythm, vocabulary choices, formality level, favorite phrases, and expressions you avoid. It then filters every piece of written output through that profile, adjusting tone for different contexts like quick Slack replies or client proposals while keeping your underlying personality consistent. The skill also enforces hard rules that remove common AI tells, including unnecessary dashes and hyphens, filler hedging phrases like “it’s worth noting” or “as you may know,” gratuitous bullet points or headers in short emails, and words such as “leverage,” “utilize,” or “delve” unless those already appear in your vocabulary. Over time the profile evolves based on your feedback, logging adjustments when you note that something is too formal or not phrased the way you would phrase it. It also learns from the style of your prompts to refine its model of how you write. Voice Mirror works with Claude Code, Cursor, Codex CLI, VS Code Copilot, and any agent that supports the SKILL.md standard.

Use cases

use it to make AI-drafted Slack replies sound like your actual quick messages

use it to remove AI tells such as unnecessary dashes, hedging phrases, and corporate filler words from generated emails

use it to train a living voice profile from your past emails, messages, and documents

use it to keep tone consistent across client proposals while adapting formality to the context

use it to refine your voice profile over time by giving feedback on generated drafts

use it to standardize written output across any SKILL.md-compatible agent or IDE

When to use it

you want AI-generated emails and messages to sound like you wrote them

you need to strip common AI tells from agent output automatically

you use Claude Code, Cursor, Codex CLI, VS Code Copilot, or another SKILL.md-compatible agent

you write in different contexts but want a consistent underlying personality

you are willing to provide writing samples and ongoing feedback to build the profile

When not to use it

you need the skill to integrate with agents that do not support the SKILL.md standard

you require support for writing languages or styles not represented in your provided samples

you need real-time collaborative editing rather than post-generation voice filtering

you cannot provide initial writing samples for the skill to learn from

you expect the skill to preserve exact formatting like dashes and bullet points that it treats as AI tells