message.md
A dynamic messaging specification for agents — built to keep AI-generated content consistent, on-brand, relevant, and uniquely yours - without breaking the token bank.
Every AI system that generates content needs a messaging house: codified positioning that defines the brand, the audience, and the portfolio. MESSAGE.md is that house — expressed as structured markdown with opinionated conventions.
Most messaging frameworks go stale the second they get written down, and agents struggle to parse messaging when multiple products, categories, personas, and alternatives are in play.
a messaging house, as code
Company facts, glossary of terms, brand guardrails, and scenario dimensions. Agents load this first, always-on.
MESSAGE.mdFoundational messaging documents — the company, narrative, market, audience, and portfolio, with the proof to back it all up.
messaging/pillars/Granular profiles loaded on demand for specific personas, products, competitors, segments, and more.
messaging/collections/Structured documents for what every asset type looks like for you, with the conventions to design and produce.
messaging/assets/the messaging room
the right message, at the right time, for the right reasons
MESSAGE.md loads first — attributes, glossary, guardrails, and the scenario vocabulary. Always on.
The agent reads the task and combines the four dimensions into the scenario at hand.
Only the pillars, collections, and assets whose Load When conditions match get pulled in.
The result is tightly scoped context assembled per task — not the entire messaging house in every prompt. Message optimized, token optimized.
tell the story
top-level positioning across the full messaging domain
hold the details
granular profiles, loaded on demand for focus and relevance
define the shape
structured output specs the agent builds against
blog-post/+ variants/customer-story/+ variants/email/+ variants/landing-page/+ variants/one-pager/+ variants/press-release/+ variants/social-post/+ variants/whitepaper/+ variants/web-copy/+ variants/slide-deck/+ variants/enablement/+ variants/LLMs already know how to write. Each asset type defines the structure and company-specific rules — variants load only when a specific one is named.