_ __ _____ _
| |/ /__ _ _ _ __ _ | ___| | _____ __
| ' // _` | | | |/ _` || |_ | |/ _ \ \ /\ / /
| . \ (_| | |_| | (_| || _| | | (_) \ V V /
|_|\_\__,_|\__, |\__,_||_| |_|\___/ \_/\_/
|___/
Agent-assisted content operations for turning documented brand strategy into reviewed drafts.
Kaya Flow helps a person or team define identity, voice, content pillars, channel rules, reusable drafting processes, and review rituals so an AI coding agent can produce better drafts without guessing who it is writing for.
Most AI content workflows fail because the agent is asked to write before the system has defined taste. Kaya Flow flips the order:
- Define the brand.
- Define the voice.
- Define the channel constraints.
- Define repeatable content processes.
- Draft, score, review, and refine.
The result is not a posting bot. It is a documented operating system for making better human-reviewed drafts.
- Brand identity and pillar templates
- Voice rules, examples, and antipattern templates
- Channel configuration templates
- Reusable content process recipes
- Agent-readable prompts for daily drafting, ad-hoc drafting, and weekly review
- Safety defaults for private data, human review, and no fabricated specifics
- Fictional examples for solo-founder and company-brand setups
Kaya Flow is draft-only by default. It does not include publisher integrations, automatic posting, account credentials, or live scheduling workflows.
The human owns the taste, facts, private context, and final approval. The agent helps turn documented source material into draft options.
git clone https://github.com/CreatorGodMode/kaya-flow.git
cd kaya-flow
mkdir -p private/brands/my-brand private/pipeline
cp -R templates/brand/* private/brands/my-brand/
cp -R templates/pipeline/* private/pipeline/Then fill in:
private/brands/my-brand/identity.mdprivate/brands/my-brand/pillars.mdprivate/brands/my-brand/voice/rules.mdprivate/brands/my-brand/voice/examples.mdprivate/brands/my-brand/voice/antipatterns.mdprivate/brands/my-brand/channels/x.mdor another channel file
Run your AI coding agent against prompts/ad-hoc.md for a single draft or
prompts/daily-loop.md for a draft batch. Review everything manually before
posting anywhere.
| Path | Purpose |
|---|---|
docs/ |
Setup, safety, voice, and dry-run workflow docs |
templates/ |
Copyable brand, channel, voice, and pipeline starters |
examples/ |
Fictional examples for learning the system |
processes/ |
Reusable content recipes |
prompts/ |
Agent-readable operating prompts |
TEXTART.md |
ASCII/TextArt marks for README, docs, and terminals |
private context -> documented voice -> process recipe -> draft -> score -> human review
A Kaya Flow run should be able to explain:
- what files it read
- what process it used
- why the draft fits the channel
- what facts need human verification
- why a weak draft was discarded
- Do not invent numbers, quotes, testimonials, events, or personal stories.
- Do not commit real brand data to a public repository.
- Do not publish automatically from this repo.
- Keep private instance data in
private/or another private repo. - Treat examples as training wheels, not source material.
Current release: v0.1.0 - public draft-only foundation.
See CHANGELOG.md for release notes.
MIT