La tua identità AI e il tuo team AI, portabili su ogni piattaforma AI.
Accedi Inizia
Menu
Crea un Agent of Me Esplora gli stili Agenti Professionali Community Agents Classifica Novità sull'AI
Piattaforme AI Directory Model Matrix Confronta Quale AI dovrei usare? Guide all'integrazione Configura OpenClaw Prompt Fit
Formazione e Strumenti Formazione Interroga i dati Agent Builder API
Info Chi siamo Contatti Disclaimer
Accedi Inizia
Account
La tua identità AI, ovunque

Crea un account gratuito per costruire il tuo profilo. Privato per impostazione predefinita. Nulla viene condiviso a meno che tu non lo pubblichi.

Inizia Accedi
Modalità scura

🧭 Vista guidata
Non conosci prompt, istruzioni di sistema, context window, token? Spieghiamo ogni termine mentre navighi, in linguaggio semplice. Stesse pagine, con la guida integrata.

⚡ Vista esperto
Sai già come funziona il prompting. Solo la sostanza, chiara e compatta, senza spiegazioni aggiuntive. Questa è la vista predefinita.

Lingua dell'interfaccia

Prompt Foundations · Sezione 4/5, Controlling the Output

Obiettivi di apprendimento
Tocca Avanti (o usa i tasti freccia) per avanzare un'idea alla volta. Nessun timer, il quiz da 5 domande ti aspetta alla fine. La ← in alto esce in qualsiasi momento; i progressi vengono salvati.

Say the shape out loud

Models default to their house style: friendly paragraphs, medium length, often more than you wanted. Every part of that is controllable, but only if you say it: 'exactly three bullets', 'one paragraph, no list', 'a two-column table: risk, mitigation', 'subject line plus five sentences'.

Think about where the output is going. Text pasted into a slide, a ticket, a contract redline and a text message have different natural shapes. Name the destination and the shape follows.

Numbers hold better than adjectives

'Short' means nothing precise; '120 words or fewer' means one thing. 'A few options' invites five; 'exactly three, one sentence each' lands. When length or count matters, use numbers, and when a hard cap matters, say it is a hard cap.

The same goes for structure: 'start with the answer in one sentence, then give reasoning' beats 'be direct'. Behavior you can count is behavior you can specify.

Define done

The strongest format instruction is a definition of done: what must be true of the output for it to be finished? 'Done means: a customer can follow these steps without asking support anything.' 'Done means: every claim has a source or is labeled an assumption.'

Definitions of done catch what templates miss, because they describe quality, not just layout. They also give you a clean way to reject: 'not done, two claims are unlabeled'.

Templates: write it once, reuse forever

If you ask for the same shape weekly, stop describing it and start providing it: a template with headers and placeholders the model fills. 'Use exactly this structure: SITUATION / OPTIONS (max 3, with trade-offs) / RECOMMENDATION / RISKS.'

Templates are the bridge from prompting to systems: they turn taste into an artifact you can reuse, share and improve, the same idea a profile applies to your identity.

Mini quiz, Controlling the Output

5 domande, estratte di volta in volta dalla banca. Soglia di superamento 60%. Tentativi illimitati.

Sezione successiva: The Iteration Loop →
Leggi il testo completo della lezione

1. Say the shape out loud

Models default to their house style: friendly paragraphs, medium length, often more than you wanted. Every part of that is controllable, but only if you say it: 'exactly three bullets', 'one paragraph, no list', 'a two-column table: risk, mitigation', 'subject line plus five sentences'.

Think about where the output is going. Text pasted into a slide, a ticket, a contract redline and a text message have different natural shapes. Name the destination and the shape follows.

2. Numbers hold better than adjectives

'Short' means nothing precise; '120 words or fewer' means one thing. 'A few options' invites five; 'exactly three, one sentence each' lands. When length or count matters, use numbers, and when a hard cap matters, say it is a hard cap.

The same goes for structure: 'start with the answer in one sentence, then give reasoning' beats 'be direct'. Behavior you can count is behavior you can specify.

3. Define done

The strongest format instruction is a definition of done: what must be true of the output for it to be finished? 'Done means: a customer can follow these steps without asking support anything.' 'Done means: every claim has a source or is labeled an assumption.'

Definitions of done catch what templates miss, because they describe quality, not just layout. They also give you a clean way to reject: 'not done, two claims are unlabeled'.

4. Templates: write it once, reuse forever

If you ask for the same shape weekly, stop describing it and start providing it: a template with headers and placeholders the model fills. 'Use exactly this structure: SITUATION / OPTIONS (max 3, with trade-offs) / RECOMMENDATION / RISKS.'

Templates are the bridge from prompting to systems: they turn taste into an artifact you can reuse, share and improve, the same idea a profile applies to your identity.

Imprese

Business AnalystChief of StaffExecutive AssistantM&A AnalystManagement ConsultantOperations AnalystProject ManagerRecruiter

Finanza

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

Legale

Contract Review AssistantLegal Due Diligence AssistantLegal Research AssistantParalegal

Marketing

Brand StrategistContent StrategistGEO AnalystMarketing StrategistSEO AnalystSales Strategist

Personale

Career CoachLearning TutorReflection AssistantResearch AssistantTravel PlannerWriting Assistant

Immobiliare

Acquisition AnalystAsset Management AnalystCommercial Real Estate AnalystDevelopment AnalystLease AnalystProperty Financial Analyst

Ricerca

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

Tecnologia

AI Strategy AdvisorCybersecurity Research AssistantData AnalystProduct ManagerSoftware Engineer