Votre identité AI et votre équipe AI, portables sur toutes les plateformes AI.
Se connecter Commencer
Menu
Créer un Agent of Me Explorer les styles Agents Professionnels Agents de la communauté Classement Actualités AI
Plateformes AI Répertoire Model Matrix Comparer Quel AI devrais-je utiliser ? Guides d'intégration Configurer OpenClaw Prompt Fit
Formation & Outils Formation Interroger les données Agent Builder API
À propos À propos de nous Contact Avertissements
Se connecter Commencer
Compte
Votre identité AI, portable

Créez un compte gratuit pour construire votre profil. Privé par défaut. Rien n'est partagé sauf si vous le publiez.

Commencer Se connecter
Mode sombre

🧭 Vue guidée
Nouveaux aux prompts, instructions système, fenêtres de contexte, tokens ? Nous expliquons chaque terme au fil de votre navigation, en langage clair. Les mêmes pages, avec l'aide intégrée.

⚡ Vue expert
Vous savez déjà comment fonctionne le prompting. Juste l'essentiel, clair et compact, sans explications superflues. C'est la vue par défaut.

Langue de l'interface

The Working Prompter · Section 1/5, Roles That Actually Work

Objectifs d'apprentissage
Appuyez sur Suivant (ou utilisez les touches fléchées) pour avancer une idée à la fois. Pas de minuteur, le quiz de 5 questions vous attend à la fin. La ← en haut permet de quitter à tout moment ; la progression est sauvegardée.

'Act as an expert' is a costume, not a brief

'You are a world-class marketing expert' rarely changes the substance of an answer. 'Expert' fits a thousand different people with a thousand different opinions, so the model plays the average of them, and the average expert sounds a lot like the assistant you already had. Stacking prestige adjectives ('brilliant, legendary, award-winning') polishes the register at best; it adds no information.

The problem is not that role-play fails. It's that the label is empty. Everything that makes a real expert useful (what they refuse to ship, who they write for, what they check first) is left for the model to guess. The fix is to stop naming the costume and start writing the brief.

The formula: role + standards + audience + stakes

A working role spells out what the title implies. 'You are a senior editor at a trade magazine. Standards: numbers over adjectives; cut anything a busy reader would skip. Audience: plant managers who know the machines but not the jargon. Stakes: this goes to print under our name.' Four short lines, and the model now has decisions to apply instead of a character to wear.

Each part does a job. The role sets the lens. The standards define what good means. The audience calibrates detail. The role says who is writing, the audience says who they write FOR. The stakes set the error tolerance: a to-print piece and a whiteboard brainstorm deserve different strictness.

Where personas pay: judgment and register

Personas earn their keep where the task has taste in it: how a careful security reviewer would push back on a design, how a support lead delivers bad news without over-apologizing, what a skeptical CFO questions first in a forecast. Judgment calls and register are exactly what a well-specified role changes.

On mechanical tasks, a persona is decoration. 'You are a meticulous paralegal' does not extract dates from a contract any better than 'extract every date, one per line, with its source sentence.' The test: if swapping in a different role would produce the same output, the role is not doing work, spend those words on the instruction instead.

Give the role opinions and a quality bar

A neutral advisor hedges: 'you could do A or B, depending.' An opinionated role decides: 'You believe most decks are twice as long as they should be; you cut ruthlessly and list what you cut.' Opinions turn a menu of options into an editorial stance, the pushback you actually wanted from an expert.

Then hand the role a quality bar it can enforce: 'reject any headline the reader couldn't repeat from memory,' 'flag every claim with no number behind it.' You stay the editor, 'keep slide 4 anyway' overrules the persona, but disagreement is now visible instead of averaged away.

Mini quiz, Roles That Actually Work

5 questions, tirées aléatoirement de la banque à chaque tentative. Note de passage : 60 %. Reprises illimitées.

Section suivante: Showing Beats Telling →
Lire le texte complet de la leçon

1. 'Act as an expert' is a costume, not a brief

'You are a world-class marketing expert' rarely changes the substance of an answer. 'Expert' fits a thousand different people with a thousand different opinions, so the model plays the average of them, and the average expert sounds a lot like the assistant you already had. Stacking prestige adjectives ('brilliant, legendary, award-winning') polishes the register at best; it adds no information.

The problem is not that role-play fails. It's that the label is empty. Everything that makes a real expert useful (what they refuse to ship, who they write for, what they check first) is left for the model to guess. The fix is to stop naming the costume and start writing the brief.

2. The formula: role + standards + audience + stakes

A working role spells out what the title implies. 'You are a senior editor at a trade magazine. Standards: numbers over adjectives; cut anything a busy reader would skip. Audience: plant managers who know the machines but not the jargon. Stakes: this goes to print under our name.' Four short lines, and the model now has decisions to apply instead of a character to wear.

Each part does a job. The role sets the lens. The standards define what good means. The audience calibrates detail. The role says who is writing, the audience says who they write FOR. The stakes set the error tolerance: a to-print piece and a whiteboard brainstorm deserve different strictness.

3. Where personas pay: judgment and register

Personas earn their keep where the task has taste in it: how a careful security reviewer would push back on a design, how a support lead delivers bad news without over-apologizing, what a skeptical CFO questions first in a forecast. Judgment calls and register are exactly what a well-specified role changes.

On mechanical tasks, a persona is decoration. 'You are a meticulous paralegal' does not extract dates from a contract any better than 'extract every date, one per line, with its source sentence.' The test: if swapping in a different role would produce the same output, the role is not doing work, spend those words on the instruction instead.

4. Give the role opinions and a quality bar

A neutral advisor hedges: 'you could do A or B, depending.' An opinionated role decides: 'You believe most decks are twice as long as they should be; you cut ruthlessly and list what you cut.' Opinions turn a menu of options into an editorial stance, the pushback you actually wanted from an expert.

Then hand the role a quality bar it can enforce: 'reject any headline the reader couldn't repeat from memory,' 'flag every claim with no number behind it.' You stay the editor, 'keep slide 4 anyway' overrules the persona, but disagreement is now visible instead of averaged away.

Entreprises

Business AnalystChief of StaffExecutive AssistantM&A AnalystManagement ConsultantOperations AnalystProject ManagerRecruiter

Finance

AccountantDue Diligence AnalystEquity Research AnalystFamily Office AnalystFinancial AnalystFixed Income AnalystInvestment Banking AnalystPortfolio Analyst

Juridique

Contract Review AssistantLegal Due Diligence AssistantLegal Research AssistantParalegal

Marketing

Brand StrategistContent StrategistGEO AnalystMarketing StrategistSEO AnalystSales Strategist

Personnel

Career CoachLearning TutorReflection AssistantResearch AssistantTravel PlannerWriting Assistant

Immobilier

Acquisition AnalystAsset Management AnalystCommercial Real Estate AnalystDevelopment AnalystLease AnalystProperty Financial Analyst

Recherche

Competitive Intelligence AnalystDeep Research AnalystIndustry Research AnalystJournalist ResearcherMarket Research AnalystMedical Research Assistant

Technologie

AI Strategy AdvisorCybersecurity Research AssistantData AnalystProduct ManagerSoftware Engineer