explaingit

prathamcreates/creative-brain

20Audience · writerComplexity · 1/5ActiveSetup · easy

TLDR

A markdown template repo a coding agent fills in by interviewing you, then uses as taste context for future creative work like writing or design.

Mindmap

mindmap
  root((creative-brain))
    Inputs
      Interview answers
      Reference links
      Feedback notes
    Outputs
      identity.md
      voice.md
      taste.md
      anti-slop.md
    Use Cases
      Brief Claude Code on your voice
      Stop AI generic output
      Capture taste references
    Tech Stack
      Markdown

Things people build with this

USE CASE 1

Have Claude Code interview you and fill in identity, voice, and taste markdown files

USE CASE 2

Feed the generated brain to Cursor before drafting blog posts or scripts

USE CASE 3

Update feedback-memory.md after each round so the agent learns what reads as off

USE CASE 4

Share a creative brain with a collaborator so they can match your tone

Tech stack

Markdown

Getting it running

Difficulty · easy Time to first run · 30min

Most of the work is the interview itself, which can take an hour or more depending on how deeply you answer the taste questions.

In plain English

Creative Brain is a small repository designed to be handed to a coding agent like Claude Code, Cursor, Hermes, or OpenClaw. You clone the repo, open it inside one of those tools, and tell the agent to use the repo to interview you and build your creative brain. The agent then asks questions about your taste, voice, references, current projects, and creative instincts, and writes the answers into a set of structured markdown files. The author's argument is that asking an AI to write or design from nothing tends to produce generic output. A creative brain is a folder of context the agent reads first, so it knows what you make, what your voice sounds like, what feels corny to you, what references shape your taste, what feels alive versus dead, and what you are working on right now. The intended audience is creatives, builders, designers, editors, writers, operators, and other internet people who want AI output that matches their taste. The repo contains a prompts folder with the agent run instructions, a templates folder with blank files the agent copies and fills in, an examples folder with sample fragments, and an outputs folder where your generated brain lives. After the interview, the agent should produce files such as identity.md, voice.md, taste.md, anti-slop.md, references.md, content-system.md, ideas.md, current-state.md, and feedback-memory.md. The README describes a feedback loop: the agent interviews you, drafts a brain, gets your feedback, updates feedback-memory.md, and uses the brain for future work. The author stresses that you should not let the agent skip the interview, since writing before understanding your taste tends to produce weak output.

Copy-paste prompts

Prompt 1
Use creative-brain to interview me about my taste and writing voice, then fill in identity.md and voice.md
Prompt 2
Run the agent through the prompts folder and produce anti-slop.md based on my answers about corny phrases
Prompt 3
Take my draft blog post and rewrite it using taste.md and references.md from my creative brain
Prompt 4
Refresh current-state.md with my active projects this month and update content-system.md to match
Prompt 5
Review the existing brain in outputs and tell me where the agent skipped the interview steps
Open on GitHub → Explain another repo

Generated 2026-05-22 · Model: sonnet-4-6 · Verify against the repo before relying on details.