← ClaudeAtlas

brand-voicelisted

Extract a brand voice from real writing samples and save it as a reusable profile. Apply saved voices to new content. No vibes — voices are learned from samples, not described in adjectives.
ImTaegan/claude-ugc-skills · ★ 1 · Data & Documents · score 74
Install: claude install-skill ImTaegan/claude-ugc-skills
# Brand Voice Most "brand voice" is vibes — "friendly, modern, approachable". That's useless. This skill extracts a *learned profile* from 5+ real writing samples, then applies it to new content. ## Instructions for Claude This skill has three modes: **extract** (build a new voice profile), **apply** (use a saved profile), **list** (show saved profiles). Detect mode from the user's input. If unclear, ask. --- ## Mode: Extract ### Step 1: Identify the Brand and Where to Save Ask: 1. **Brand name** (used as the filename, e.g., "mintlooks" → `mintlooks.md`) 2. **Save location**: project (`.claude/brand-voices/`) or global (`~/.claude/brand-voices/`)? ### Step 2: Collect Samples Ask for AT LEAST 5 writing samples. More is better. Tell the user: ``` I need at least 5 samples to extract a voice. The more diverse the better: - LinkedIn posts - Instagram captions - Newsletter excerpts - Brand guidelines - Customer-facing emails - Existing video scripts Paste them here, one per message or all in one — your call. Tell me when you're done. ``` If the user provides fewer than 5, push back: ``` That's [N] samples — fewer than I'd want for a reliable profile. With this few I'll guess more than I learn. Either: 1. Send more samples (recommended) 2. Continue with what we have, knowing the profile will be rougher ``` ### Step 3: Extract the Profile Read all samples carefully. Extract: **Tone signature** (3–5 words): Look at HOW the brand says things, not what they say. Example