10 blog ideas. Same model, same account, two minutes apart.
The first prompt was one line: give me 10 blog post ideas for my bakery. It returned ideas that fit any bakery on earth.
The second was the same request carrying 15 components: who reads it, what they buy on a Saturday, the tone, the format, five things to avoid, and a flag on any claim the model cannot verify.
Same model. Same bakery. One variable changed: the brief.
I spent 15 years telling junior writers that a vague brief returns vague copy. Language models are stricter than any writer I have trained. They fill every gap you leave with the most average answer they hold.
So I wrote down the 15 parts I check before any prompt that matters — plus a meta-prompt that runs the check for you. Paste your rough prompt at the bottom, get back a rebuilt one: role, audience, quality bar, and a [CHECK] flag on anything unverified.
13 pages, free, no card details. It is the guide I wish I had handed my team in 2023.
Which component do your prompts skip most often? Mine was stakes, for about a year, and the output stayed polite and forgettable the whole time.
📄 13 pages + the prompt as a text file.
🆓 Free to use.
Works with ChatGPT, Claude, and Gemini.
