Voice Lab

Voice Lab

Build a voice profile for this repo's content pipeline.

This is not a generic brand exercise. The output should help downstream skills write hooks, angles, and concept bodies that sound like the creator, not like an AI summary.

Output File

Save the final profile to:

~/clawd/workspace/brand/voice-profile.md

If useful observations come up during the process, append them to:

~/clawd/workspace/brand/notes.md

Modes

1. Extract

Use when the user provides 3 or more writing samples.

Goal:

• infer the creator's voice from real material
• capture repeatable patterns, not vague adjectives

2. Build

Use when the user has little or no source material.

Goal:

• ask focused questions
• build a practical house voice for this project

3. Refine

Use when voice-profile.md already exists.

Goal:

• preserve what still works
• tighten weak sections
• add newly observed phrases, constraints, and platform preferences

Workflow

Step 1: Check for an existing profile

If ~/clawd/workspace/brand/voice-profile.md exists:

• summarize what it currently says
• ask whether to refine it or replace it

Step 2: Gather inputs

For Extract mode, collect:

• writing samples
• audience description
• topics they cover often
• topics or tones they want to avoid

For Build mode, ask about:

• audience
• expertise
• tone
• level of directness
• willingness to be contrarian
• proof style
• taboo phrases or styles

Step 3: Produce a practical profile

The profile must be useful to content-dna and the-forge.

Include these sections:

# Voice Profile

## Summary

## Audience

## Tone Sliders
- Direct vs soft
- Tactical vs inspirational
- Contrarian vs consensus
- Polished vs conversational

## Writing Rules

## Vocabulary To Lean Into

## Vocabulary To Avoid

## Signature Moves

## Hook Patterns

## Proof And Specificity Rules

## Platform Notes
- X
- LinkedIn
- Short video
- Newsletter

## No-Go List

## Sample Before/After Rewrites

Step 4: Validate before save

Generate three short samples in the proposed voice:

• an X opener
• a LinkedIn opening paragraph
• a short video hook

Ask the user if the voice feels right.

Step 5: Save

Save the final version to ~/clawd/workspace/brand/voice-profile.md.

If the user wants to keep evolving it, append extra observations to ~/clawd/workspace/brand/notes.md.

Rules

• Prefer concrete patterns over personality fluff.
• Capture rhythm, proof style, and sentence habits.
• Do not overwrite an existing profile silently.
• If samples conflict, say where the voice is mixed.
• Keep the profile compact enough for downstream prompting.