Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/LilMGenius/polysona/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks you through cloning Polysona, connecting it to your AI agent runtime of choice, and completing your first psychology interview. By the end, you will have a working persona extracted and ready to drive the content pipeline. The whole process takes under five minutes.

Prerequisites

Before you begin, make sure you have the following available:
  • Node.js (v18 or later) or Bun (v1.0 or later) installed on your machine
  • Codex (OpenAI) or Claude Code (Anthropic) available in your development environment
  • A terminal open and ready

Installation and Setup

Codex is the primary integration target for Polysona. AGENTS.md is auto-recognized by Codex, and repo skills are auto-discovered from .agents/skills.
1

Clone the repository

Clone Polysona and navigate into the project directory.
git clone https://github.com/LilMGenius/polysona.git && cd polysona
2

Install dependencies

Install all required packages using Bun.
bun install
3

Sync Codex skills

This repo mirrors ./skills into ./.agents/skills for Codex compatibility. Run the sync script to make sure the latest skills are available.
bun run codex:skills:sync
Alternatively, you can run the script directly:
node ./scripts/sync-codex-skills.mjs
If an updated skill does not appear immediately in Codex after syncing, restart your Codex session. Codex caches skill definitions at session start.
4

Start a Codex session

Open a Codex session in the polysona repo directory. Codex will automatically detect and load AGENTS.md, making all persona agents and skills available immediately.
5

Run your first interview

In your Codex session, trigger the profiler agent to begin the deep psychology interview.
$interview
The profiler will guide you through a structured interview using 10 psychology frameworks. When complete, Polysona structures the results into persona.md, nuance.md, and accounts.md under personas/default/.

Running the Full Pipeline

After completing the interview, you have a structured persona. The full content pipeline flows through five stages in sequence.
$interview → $introduce → $trend → $content [platform] → $qa → $publish
Run $introduce (or /introduce) at the start of any new session to inject the active persona into the agent context before generating content. Then follow the pipeline in order. Example: generating a LinkedIn post
$introduce
$trend
$content linkedin
$qa
$publish

Command Reference

The following commands are available in both Codex ($command) and Claude Code (/command) sessions.
CommandPurpose
$interview / /interviewStart a deep psychology interview using 10 frameworks (profiler agent)
$introduce / /introduceInject the current active persona into the session context
$trend / /trendScan your domain for trending topics filtered to your persona (trendsetter agent)
$content [platform] / /content [platform]Generate persona-conditioned drafts for the specified platform (content-writer agent)
$qa / /qaSimulate follower and rolemodel reactions, score drafts, and surface the top 5 (virtual-follower agent)
$publish / /publishPublish the selected draft and capture engagement metadata (admin agent)
$status / /statusDisplay the active persona summary and current pipeline status
$export / /exportExport persona-derived instructions into CLAUDE.md and AGENTS.md for use in another workspace
The [platform] argument for $content accepts: x, threads, linkedin, naver, or brunch. Always specify the platform so the content-writer can apply the correct format, length, and tone patterns.

Next Steps

Once your persona is set up and you have run your first pipeline cycle, explore these resources to go deeper.
  • Architecture — Understand how the orchestrator, agents, skills, and data layer fit together.
  • Agents — Read the full specification for each of the five agents.
  • Skills — Explore the 8 skills and how they are stored and loaded per runtime.

Build docs developers (and LLMs) love