| name | on-brand-content |
| description | Produce marketing content that stays strictly on-brand: apply a brand's positioning, voice/tone, audience, do's/don'ts, and guardrails from AGENTS.md/SOUL.md before drafting or revising any copy. Use whenever writing social posts, captions, ad copy, emails, or briefs for a specific brand. |
| metadata | {"hermes":{"tags":["marketing","brand-voice","content","kaizen"],"homepage":"internal"}} |
On-Brand Content
Guidance for turning a content brief into copy that reads like it was
written by the brand itself, not a generic AI assistant. This is the shared
playbook for Kaizen's Content Creator and Brand Strategist specialists.
Before writing a single word
- Read the brand DNA first. Every tenant has brand identity in
SOUL.md (agency identity + hard guardrails) and AGENTS.md (editable
brand DNA: positioning, voice/tone, audience, do's/don'ts, channels).
Never draft content without reading both — a brief alone is not enough
context.
- Identify the channel. Tone, length, and structure differ by channel
(e.g. Twitter/X thread vs. LinkedIn post vs. email vs. blog). If the
brief doesn't specify a channel, default to the brand's primary channel
from
AGENTS.md, and say so explicitly in your output.
- Check the guardrails before the voice. Guardrails (e.g. "never
promise a delivery date we haven't confirmed") are hard constraints —
violating one is worse than an off-tone draft. Voice/tone notes are
stylistic; guardrails are non-negotiable.
Applying voice and tone
- Match the brand's documented tone adjectives literally (e.g. "confident,
plain-spoken" means short declarative sentences, not hedging language or
corporate throat-clearing).
- Mirror the brand's documented do's: if "use concrete numbers" is a do,
every draft should contain at least one concrete, specific detail (a
number, a named feature, a real outcome) rather than vague superlatives.
- Respect documented don'ts literally (e.g. "don't use exclamation points"
means zero
! in the draft, not "use them sparingly").
- When positioning and audience conflict with a generic best-practice
(e.g. "always add a CTA"), the brand's own documented preferences win.
Avoiding generic AI-slop copy
Do not ship copy that could be mistaken for a random SaaS company's
content. Concretely avoid, unless the brand's own voice explicitly calls
for it:
- Opening with "In today's fast-paced world..." or similar throat-clearing
- Rhetorical questions as a crutch ("Ever wondered how...?")
- Emoji stacking or exclamation-heavy enthusiasm
- Vague superlatives ("game-changing", "revolutionary", "seamless") without
a concrete claim backing them up
- A generic three-point listicle structure applied to everything regardless
of the actual brief
Instead: lead with the single most concrete, specific claim from the brief,
support it with one real detail (a number, a named capability, a specific
outcome), and close with a call to action that matches the brand's
documented voice — not a boilerplate one.
Research before claims (when Linkup or web tools are available)
If a claim in the brief is unverified (a competitor comparison, a market
stat, a "first/only" claim), verify it with the research tools available
(Linkup search when configured, otherwise web_search/web_extract)
before stating it as fact. Never fabricate a statistic, quote, or
competitor detail — cite the source inline as a note if the content format
allows it, or flag the unverified claim explicitly in your final response
if it doesn't.
Before finalizing
Run through this checklist on every draft:
If any box can't be checked, say so explicitly in the response rather than
shipping a draft that silently fails the brand's own standards.
