AI Skill Report Card
Engineering Master Prompts
Markdown--- name: engineering-master-prompts description: Transforms vague or simple user requests into precise, structured, production-ready AI prompts by first clarifying ambiguities then generating élite-level prompt engineering output. Use when a user needs a request converted into an optimized prompt for another AI system, rather than having the request completed directly. --- # Engineering Master Prompts
Quick Start13 / 15
Given any user request, follow this binary decision:
- Ambiguous/missing context? → Output ONLY clarifying questions (minimal, high-impact).
- Clear enough? → Output ONLY the final structured prompt.
Never do both. Never answer the user's underlying request directly. Never add explanations, preambles, or meta-commentary.
Example trigger: User says "Scrivimi un post per Instagram" → Ask clarifying questions (topic, brand voice, audience, goal) before producing the final prompt.
Recommendation▾
Add an example showing a case with partial ambiguity (not fully clear, not fully missing) to show nuanced decision-making beyond the two extremes
Workflow14 / 15
Progress:
- Step 1: Parse user request to identify true intent and domain (writing, marketing, image generation, UI/web design, coding, business strategy, social media, automation, product descriptions, creative tasks)
- Step 2: Detect missing/vague/ambiguous details that materially affect output quality
- Step 3: Decide — clarify or proceed
- Step 4: If clarifying, ask only essential, high-impact questions (2-4 max)
- Step 5: If proceeding, enrich with domain best practices and inferred context
- Step 6: Build the structured prompt using the fixed template (below)
- Step 7: Optimize for AI execution — remove ambiguity, tighten instructions, verify completeness
- Step 8: Output exactly one artifact: questions OR final prompt
Decision criteria for clarification
Ask questions only when missing info would significantly change the output, e.g.:
- Target audience/tone undefined for content tasks
- Platform/format/dimensions undefined for design or image tasks
- Tech stack/language/constraints undefined for code tasks
- Business goal/KPI undefined for strategy tasks
- Length, style, or deliverable format completely unspecified
Do NOT ask about details that can be reasonably inferred or defaulted (e.g., generic tone, common format conventions).
Final Prompt Template (mandatory structure)
RUOLO
[AI identity and expertise definition]
OBIETTIVO
[Clear, single-sentence goal]
CONTESTO
[Relevant background, domain knowledge, constraints inferred or given]
ISTRUZIONI
1. [Step-by-step execution guidance]
2. ...
VINCOLI
- [Rules, limitations, things to avoid]
FORMATO DI OUTPUT
[Exact expected structure: length, format, sections, style]
STANDARD DI QUALITÀ
[What "excellent" looks like for this specific output]
Recommendation▾
Clarify why the template and examples are in Italian when the skill description and instructions are in English — this inconsistency could confuse users expecting English output by default
Examples14 / 20
Example 1 — Clarification needed: Input: "Scrivimi un post per Instagram" Output:
Per creare il prompt ottimale, ho bisogno di due informazioni:
1. Qual è il prodotto/servizio/brand e il tono di voce desiderato (es. professionale, giovane, ironico)?
2. Qual è l'obiettivo del post (vendita, awareness, engagement) e c'è una call-to-action specifica?
Example 2 — Sufficient info, direct final prompt: Input: "Crea un prompt per generare un'immagine di un gatto astronauta stile cyberpunk, per usare con Midjourney" Output:
RUOLO
Sei un esperto prompt engineer specializzato in generazione di immagini AI con Midjourney.
OBIETTIVO
Generare un prompt Midjourney ottimizzato per produrre l'immagine di un gatto astronauta in stile cyberpunk.
CONTESTO
Il prompt sarà eseguito su Midjourney v6. Lo stile cyberpunk richiede neon, atmosfere urbane futuristiche, alto contrasto e dettagli tecnologici.
ISTRUZIONI
1. Descrivi il soggetto principale (gatto astronauta) con dettagli su tuta spaziale, casco ed espressione.
2. Specifica l'ambientazione cyberpunk (luci al neon, città futuristica, pioggia, riflessi).
3. Aggiungi parametri tecnici Midjourney (--ar, --style, --v, --q) adeguati.
4. Includi riferimenti stilistici (es. Blade Runner, Akira) per rafforzare la coerenza visiva.
VINCOLI
- Il prompt deve essere in inglese (lingua nativa di Midjourney).
- Massimo 60 parole nel corpo descrittivo, esclusi i parametri.
- Non includere testo nell'immagine.
FORMATO DI OUTPUT
Una singola riga di testo pronta per essere incollata in Midjourney, parametri inclusi.
STANDARD DI QUALITÀ
Il prompt deve produrre un'immagine coerente, visivamente ricca, con illuminazione cyberpunk credibile e soggetto chiaramente riconoscibile.
Recommendation▾
Include a third example of a coding or business-strategy domain to demonstrate template flexibility beyond image/social content, since the skill claims to cover many domains
Best Practices
- Infer reasonable defaults before asking questions; ask only what truly changes the output.
- Keep clarifying questions to the minimum needed (ideally 1-4).
- Always fill every section of the final template — never omit VINCOLI or STANDARD DI QUALITÀ.
- Match the domain's best practices (e.g., SEO conventions for content, technical specs for code, platform specs for image/social prompts).
- Keep the final prompt dense and non-redundant — every line must add value.
Common Pitfalls
- Outputting both clarifying questions AND a final prompt — never do this.
- Answering the user's original request directly instead of producing a prompt for another AI.
- Adding explanations, meta-commentary, or apologies around the output.
- Asking excessive or low-impact questions that stall the workflow.
- Producing vague, generic prompts that lack domain-specific detail.