How this works

Guide

A quick tour of the workspace. Anything more specific, just ask the co-pilot.

Overview

Counts at a glance. Recently touched files. The complete script surface this rig can run.

Brands

One card per brand. Click in for the style guide, deliverables, and analytics presence.

Agents

Each agent has a Skill (prompt) and Learnings (memory). The co-pilot can read or refine both.

Actions

A curated allowlist of project verbs: Pull GA4, fetch Google Ads, run a tracking audit, refresh team capacity, and so on. Forms in, jobs out.

Jobs

Every action lands here as a tracked job with a live log.

Processes

The few long-running services: activity tracker, Slack bot bridge, Shopify proxy. Start, stop, tail.

The Co-pilot

Press / from anywhere. Ask in natural language. Examples:

  • · "Run a CRO audit for <brand>"
  • · "Show me the copywriter skill prompt"
  • · "Refine the brand-extractor agent: detect sticky headers"
  • · "Pull the last 30 days of Google Ads for customer 123-456-7890"
  • · "Which brands have no learnings file yet?"

What the co-pilot can do

  • · Read every brand guide, learning, and workflow doc.
  • · Read every output file in outputs/.
  • · Edit skill prompts, learnings, brand guides, and outputs.
  • · Trigger any allowlisted Action.
  • · Create new files inside the project's content folders.

What it can never do

  • · Read .env, credentials, or service-account keys.
  • · Print or expose API keys, even when scripts use them.
  • · Touch its own dashboard implementation.
  • · Move, rename, or delete folders.
  • · Run network installs (curl, npm install, etc.).