Test final: The Working Prompter
Dix questions couvrant toutes les sections, une sélection fraîche et équilibrée de l'ensemble de la banque à chaque tentative. Réussite à 60 % ; un nouvel essai propose toujours des questions différentes.
Test final
10 questions, tirées aléatoirement de la banque à chaque tentative. Note de passage : 60 %. Reprises illimitées.
Capstone: The Reusable Toolkit
Pick a task you repeat at least weekly. Build a prompt template for it: labeled sections, {{variables}} for everything that changes per use, instructions cleanly walled off from pasted material, and a definition of done. Add a style block, three to six concrete voice rules (at least one negative) plus one 100-200 word sample of your real writing, and one labeled example of a great output. Then run the assembled toolkit on a real instance of the task.
Barème: Template 40: labeled sections with clear delimiters (10), variables cover exactly what changes per use (10), instructions separated from material (10), a definition of done (10). Style block 30: rules concrete and testable (10), at least one negative rule naming a real tic (10), sample genuinely your writing in the target medium (10). Labeled example 20: matches the desired format exactly (10), labeled so it can't be mistaken for material (10). Road test 10: you ran it on a real instance and recorded one fix.
Avant de commencer: Choose the recurring task before you start, and dig up two real past outputs you were happy with, one becomes your labeled example, the other your style sample.
Le projet final est auto-évalué selon la grille. C'est ce que vous conservez : un vrai livrable, construit avec tout ce que le cours a couvert.