Own practice · two brands
An editorial system with enforced brand voice
Long-form plus channel teasers from a single run, with brand separation enforced by the skill rather than by review.
Stack
- Claude Skills
- Brand voice specification
- Multi-channel derivation
The problem
Publishing consistently across a newsletter, a professional network and social channels, for two brands with deliberately different audiences, voices and calls to action.
Generic AI output was worse than useless here: it reads like everyone else’s, and it mixes the brands.
The approach
A content skill per brand that encodes the voice rules explicitly: sentence structure, paragraph length, what emphasis is allowed, which call to action belongs to which brand, and what is forbidden outright.
One run produces the long-form article plus the platform-native teasers derived from it, not reformatted copies of it.
The result
Long-form plus channel teasers from a single run, in the correct voice, with brand separation enforced by the skill rather than by review.
The transferable piece
Voice is a specification, not a prompt. Written down properly, it survives being run a hundred times.