Five fields, and the prompt writes itself in front of you. Context, role, action, format, tone: you fill them in, it assembles, you copy it into your tool. The same workshop audits a prompt you already wrote.
242 prompts drawn from the lessons. Sorted by module, copied in one gesture, opened directly in Claude or ChatGPT. Your own get saved there too.
Real work situations, not demo examples. Minutes to anonymise, an ad to decline in variants, an export to clean: the request, the answer, and what made the difference.
Submitted from the lesson, on your own data. One quiz per module, graded server-side. Exercises are reviewed by Paul with Automations, 3 reviews included.
It answers from your lessons, and cites the source. A question at 11 pm, an answer drawn from the programme, never invented: it only knows what you have.
Agents, automations, MCP servers. 73 lessons in their own course and seventeen real cases from the agency. This is where you wire the tools to one another.