Brand Voice skill
The playbook Moonrank's Writer Agent uses to turn "friendly but professional" into a real voice document — concrete attributes, paired examples, and rules that settle edit debates.
Download the methodology, see how the agent extracts a voice from your best pieces, or run it live inside the Writer Agent.
- Category
- Writer
- Agent
- Writer Agent
- Format
- brand-voice.md
- Price
- Free to download
The skill file
What is the Brand Voice skill?
The Brand Voice skill is how Moonrank's Writer Agent pins down what your brand sounds like: extracting attributes from pieces that already sound right, pairing each one with on-voice and off-voice examples, and writing the mechanics and banned moves down where a writer can use them.
It's written as a SKILL.md, a plain-language instruction format that people and AI agents can both follow. Read it to build your own voice document, hand it to freelancers as the tiebreaker in edit debates, or let the Writer Agent derive the voice from your published work and write against it.
Downloading the playbook is free with a Moonrank account. Applying the voice to every draft, every day, without drift is what the Writer Agent does.
See it in action
“Our style guide says "confident, approachable, expert" and every writer interprets it differently.”
Those three adjectives describe 90% of B2B blogs — no writer can be wrong, so no writer can be right. From your 12 best-performing posts, the actual voice is specific: second person, sentences averaging 14 words, one dry joke per section, zero exclamation marks. The rewrite is a document with 4 "X, not Y" attributes, 8 paired examples, and a banned list of 15 quoted phrases. Edit debates end because the document answers them.
From input to fix list
- 1
You provide the evidence
Three to five pieces that sound right, one or two that don't — the voice lives in the gap.
- 2
Moonrank extracts the attributes
The Writer Agent names what your best sentences do, as "X, not Y" boundaries rather than adjectives.
- 3
It writes the paired examples
One on-voice and one off-voice sentence per attribute, on the same subject — the pair is the rule.
- 4
It documents mechanics and bans
Person, sentence length, jargon and humor policy, plus the exact phrases that never appear.
Everything it checks
Attribute set
Three to four voice attributes, each with a boundary — "direct, not blunt", never "authentic".
Paired examples
On-voice and off-voice sentences on the same subject, one pair per attribute.
Voice mechanics
Person, contractions, sentence length range, jargon policy, humor policy.
Banned list
The openers, hedges, and clichés that mark a draft as off-brand — quoted literally.
Edge-case guidance
How the voice flexes for bad news, legal copy, and support replies.
Convergence test
A pass/fail check: two people using the document should produce the same rewrite.
When to reach for it
- End edit debates that are really taste debates.
- Onboard freelancers who sound like you by their second draft.
- Give an AI agent a voice it can hold at publishing volume.
- Turn a vague style guide into rules that settle arguments.
- Keep one voice across blog, email, and social as the team grows.