For developers and content professionals managing communication for multiple voices, consistency and authenticity are critical. We're introducing draft-in-voice, an AI agent skill designed to help with this specific challenge.
draft-in-voice generates 2-3 written options in a specific person's voice. This covers various content types like tweets, posts, emails, and launch copy. The core utility is its ability to apply an existing, validated voice profile, rather than inventing one. This ensures that the output aligns with a known communication style.
The tool is particularly useful for anyone needing to draft communications that genuinely sound like a specific individual, where maintaining that personal style is key to effective messaging. It's not a general AI humanizing, summarization, or corpus-building platform, but a focused skill for voice-specific content generation.
How the Tool Operates
When a request is made, draft-in-voice follows a clear process. First, it loads the designated person's voice profile, identified by its people/<slug>-voice identifier stored in the system's brain. Following this, it pulls relevant contextual information from the brain to ensure the draft is informed and appropriate.
After gathering context, it drafts several options in the right register. These options are not immediately presented. Instead, the tool runs a rigorous self-check against predefined parameters. These include named tells, length discipline, register purity, and factual accuracy. Only after passing this internal validation are the results shown; nothing is auto-posted.
This methodical approach ensures that the generated drafts are not only stylistically accurate but also adhere to specific communication guidelines before ever reaching a human for review.
Interacting with the Skill
Users can trigger draft-in-voice with a few direct commands. These include phrases like 'draft in voice', 'write this as', 'ghostwrite', 'in their voice', or 'build a voice profile'. Each trigger initiates the process of generating voice-specific content.
It's important to note a key requirement: the tool needs a pre-validated voice profile to function. Without such a profile, it will refuse to draft content. This strict condition prevents the generation of unverified or generic voice outputs, reinforcing its commitment to authenticity.
Again, to be clear on its scope, draft-in-voice is not intended for general AI humanizing, summarizing lengthy texts, or building large text corpuses. Its purpose is narrowly defined: to produce specific written options in a known individual's established voice.
The Voice Self-Check Explained
The internal self-check is where draft-in-voice truly validates each draft. This process runs against six hard rules to ensure quality and authenticity. First, it ensures the draft avoids any signature tells that are specifically known to be out of character for the profile.
Second, it checks that the draft matches the median length typically associated with that person's communications for a given format. Third, it verifies the register purity, ensuring the tone and style are consistently accurate. Fourth, it respects banned phrases identified in the profile, making sure they are not used.
Fifth, the tool assesses if the draft sounds authentically like the individual. Finally, it traces the substance of the content back to documented sources to ensure factual accuracy. This comprehensive validation process is critical.
For example, imagine needing to draft a brief internal memo from a senior manager. The skill would generate a few versions. During its self-check, it would confirm if the language matches the manager's typical formal but concise style, avoids any specific idioms they never use, and adheres to their usual memo length. If the manager consistently uses short, direct sentences, the tool will ensure the drafts reflect that, providing options that truly sound like they came from that manager.
Frequently Asked Questions
Q: What kind of content can draft-in-voice create?
A: It can generate tweets, posts, emails, and launch copy in a specific person's voice.
Q: Is content automatically published?
A: No, nothing is auto-posted. The tool generates options for review.
Q: What does it mean by a 'validated voice profile'?
A: draft-in-voice requires a pre-validated profile for a specific person before it can draft content.
This focused AI agent skill provides a practical way to ensure consistent, authentic communication for specific individuals. It streamlines the drafting process while upholding strict standards for voice accuracy and factual integrity.
