Draft a Prompt

Interactive Field Manual & Reference
← Return to Prompt Workspace
📖 ENGINEERING SPECIFICATION & MASTERCLASS

The Draft a Prompt Field Manual

Why basic AI prompts fail, how the RTC-CO framework delivers 10x signal, an interactive directory of all 12 specialized domain engines, and the exact decision matrix for API keys.

🧠 The 12 Built-In Domain Engines

Every engine is engineered with deep domain expertise. Click Launch in Workspace on any card to automatically pre-populate the Draft a Prompt workspace with that engine's configuration.

🔍
📄

1. Resume & Career Strategy

Career

ATS keyword alignment, Google XYZ impact statements, and executive career narratives.

Persona:Elite Senior Technical Recruiter & Career Strategist
Guardrails:Zero passive filler; strictly enforce Google XYZ impact formula
✨ High-Impact Output Sample
"Architected distributed PySpark ETL pipelines on AWS EMR processing 4.2TB daily, reducing ingestion latency by 43% and saving $38K/quarter in cloud compute."
⚡ Launch in Workspace →
✉️

2. Cold Email & Sales Cadence

Sales

High-deliverability 3-touch B2B outbound sequences engineered for busy technical executives.

Persona:Elite B2B Outbound Demand-Gen & Direct Copywriting Director
Guardrails:Under 85 words/email; low-friction relevance CTA; zero generic pleasantries
✨ High-Conversion CTA Sample
"Saw your team migrated EMEA to Snowflake. Most teams see query costs surge 3x by month 2. Worth exploring if this matches your Q3 cost-containment roadmap?"
⚡ Launch in Workspace →
📊

3. Executive Summary & Brief

Leadership

BLUF (Bottom Line Up Front), high signal-to-noise ratio, risk matrices, and capital allocation decisions.

Persona:Chief of Staff & Senior Management Consultant to Fortune 100 C-Suite
Guardrails:Strict 1-page density limit; quantify all ROI & risks; binary decision path
✨ Authoritative BLUF Sample
"Bottom Line Up Front: We recommend approving Option B (Managed Cloud Migration at $240K CapEx), eliminating our on-prem single-point-of-failure before peak Q4 volume with an 11-month payback."
⚡ Launch in Workspace →
📈

4. Data Engineering & Analytics

Data

PySpark/SQL transformations, multi-terabyte schema architecture, data quality assertions, and statistical diagnostics.

Persona:Staff Machine Learning Engineer & Principal Distributed Data Architect
Guardrails:Idempotent MERGE ON writes; schema evolution; automated null/drift quarantine
✨ Data Quality Contract Sample
"1. Null Check: Route null user_id to /quarantine/events. 2. Drift: Alert if p99 session duration diverges > 2.5 std-devs from 30-day baseline. 3. Idempotency: MERGE ON (event_id, date_partition)."
⚡ Launch in Workspace →
💻

5. Software & System Arch

Engineering

Clean Architecture, SOLID principles, API contracts, defensive error boundaries, and microservice resilience.

Persona:Principal Software Architect & Systems Engineer (15+ yrs distributed scale)
Guardrails:Explicit failure modes; circuit breakers; time/space complexity trade-offs
✨ Resilient Interface Sample
"interface OrderService { createOrder(ctx: Context, req: CreateOrderRequest): Promise>; } Failure Mitigation: If payment times out >3000ms, trip breaker & emit dead-letter queue event."
⚡ Launch in Workspace →
🏠

6. Real Estate & PropTech

PropTech

RESO Web API compliance, real-time MLS syndication, geospatial polygon queries, and valuation engines.

Persona:Principal PropTech Architect & Real Estate Platform Strategist
Guardrails:Strict RESO Data Dictionary compliance; PostGIS bounding-box indexing
✨ RESO Query & Spatial Sample
"GET /Property?$filter=ModificationTimestamp gt 2026-10-19T00:00:00Z&$top=500 SELECT id, price, ST_AsGeoJSON(geom) FROM listings WHERE ST_DWithin(geom, ST_MakePoint(:lng,:lat)::geography, :radius) LIMIT 50;"
⚡ Launch in Workspace →
🔄

7. Two-Sided Marketplaces

Platform

Cold-start liquidity, supply/demand matching engines, Stripe Connect split-escrow, and take-rate economics.

Persona:Platform Economist & Marketplace Systems Architect
Guardrails:Split-payment milestone release; anti-disintermediation communication boundaries
✨ Escrow Protocol Sample
"1. Buyer checkout authorizes charge (capture_method=manual). 2. Provider fulfills -> Buyer verifies completion. 3. System captures funds, auto-splits 85% to Provider and 15% to platform treasury balance."
⚡ Launch in Workspace →
✍️

8. Content & Copywriting

Creative

High-CTR hooks, scannable platform-native formatting, viral distribution mechanics, and zero corporate fluff.

Persona:Senior Content Strategist & Direct-Response Editorial Director
Guardrails:Scroll-stopping hook in line 1; 1-2 sentence paragraphs; ban AI tropes ("delve", "tapestry")
✨ 3 Hook Archetypes Sample
"1. Contrarian: '90% of teams build microservices for scale. In reality, it slows delivery by 4x for teams under 50 devs.' 2. Data: 'We analyzed 1,200 slow queries. 78% were caused by one missing compound index.'"
⚡ Launch in Workspace →
🚀

9. Business Strategy & Pitch

Strategy

Bottom-up TAM/SAM/SOM calculations, defensible competitive moats, unit economics (CAC, LTV, payback), and investor theses.

Persona:Former Tier-1 Venture Partner & Principal Strategy Consultant (McKinsey/Bain)
Guardrails:Rigorous bottom-up sizing; verifiable customer segments; concrete CAC/LTV math
✨ Bottom-Up Sizing Sample
"TAM: $14.2B (120K mid-market logistics firms x $118K ACV). SAM: $3.8B (32K cold-chain carriers). SOM: $190M (1,600 carriers over 36 mos). Unit Economics: CAC = $9.2K, ACV = $48K, LTV = $168K (18:1 ratio), Payback = 7.1 mos."
⚡ Launch in Workspace →
🎓

10. Education & Learning

Education

Feynman technique, cognitive load management, intuitive physical analogies, and first-principles deconstruction.

Persona:Master STEM Educator & Technical Curriculum Architect
Guardrails:Explain 'why invented' before 'how it works'; minimum 2 physical analogies
✨ Physical Feynman Analogy Sample
"Public-Key Crypto Analogy: Alice leaves an open padlock (Public Key) on a wooden box in the square. Anyone can drop a note in and click it shut. But only Alice holds the brass key (Private Key) that can unlock it."
⚡ Launch in Workspace →
👥

11. HR, Recruiting & Rubrics

People

Outcome-based 30/60/90 scorecards, bias-free competency matrices, and structured STAR interview evaluation rubrics.

Persona:VP of People Operations & Organizational Design Specialist
Guardrails:Separate must-haves from nice-to-haves; objective 1-4 anchors prevent interviewer bias
✨ Behavioral Rubric Anchor Sample
"Question: 'Describe a project at risk due to misalignment.' Level 1 (Fail): Blamed other team; escalated passively. Level 4 (Superior): Identified structural gap, authored decision matrix, met deadline & created permanent cross-team SLA."
⚡ Launch in Workspace →
📱

12. UX/UI & Product PRDs

Product

Comprehensive PRDs, Gherkin acceptance criteria (Given/When/Then), 4 UX states (Empty/Loading/Error/Success), and WCAG accessibility.

Persona:Staff Product Manager & Principal UX Systems Designer
Guardrails:Explicit Non-Goals; Gherkin scenarios; WCAG 2.1 AA accessibility mandates
✨ Gherkin Acceptance Criteria Sample
"Scenario: Expired card checkout Given 3 items in cart & saved card is expired When user clicks 'Complete Order' Then show inline error 'Card expired' And do NOT decrement inventory And preserve cart selections."
⚡ Launch in Workspace →

⚡ The RTC-CO Framework Blueprint

Why conversational prompts fail and how to treat your AI prompt like a production engineering specification.

Large Language Models are probabilistic next-token prediction engines. When fed ambiguous instructions like "write a marketing email" or "help me optimize this database", the model is forced to infer dozens of missing variables. It defaults to the statistically average response: generic corporate buzzwords, conversational preambles, and ungrounded claims.

R
Role & Persona
Assign exact seniority, credentials, and authority. Saying "Act as an expert" yields generic advice. Saying "Principal DevSecOps Architect with 15+ years enterprise banking experience" primes the model's highest-signal weights.
T
Task (Objective)
Define a single, ranked primary deliverable. Avoid kitchen-sink tasks where the model must do 10 unranked actions at once. Sequence secondary instructions logically.
C
Context (Environment)
State the exact target audience, tech stack, scale, and operational realities. Context stops the AI from assuming standard SaaS defaults when you are actually in high-security banking or healthcare.
C
Constraints (Guardrails)
Negative boundaries are the secret weapon of prompt engineering. Mandate zero conversational preambles, ban filler buzzwords, enforce Google's XYZ formula, and set strict word ceilings.
O
Output Format
Mandate exact headers, Markdown tables, or JSON schemas. When output structure is declared, the model spends zero compute guessing how to present information.
E
Examples (Few-Shot)
Show, don't just tell. Providing even 1 concrete input-output demonstration or counter-example calibrates tone, syntax, and density more effectively than 500 words of instructions.

🔑 How & When to Use API Keys

Understanding the difference between the Built-In Local Privacy Engine and Frontier Cloud Models.

Feature Built-In Local Privacy Engine (Default) Frontier Cloud Models (Optional API Key)
Cost 100% Free Forever Pay-per-token via your own provider account (Google AI Studio has free tier)
Execution Speed Instant (< 1 ms) client-side execution 1–4 seconds cloud network latency
Privacy & Data Flow 100% Client-Side. Zero server transmission. Browser communicates directly with Google, OpenAI, Anthropic, or DeepSeek
Setup Required Zero setup. Works immediately out of the box. Requires generating an API key from provider developer console
Best Suited For 90% of prompt drafting, template execution, manual tuning, and learning Open-ended creative generation from rough notes using frontier LLMs

Official Developer Key Consoles:

If you choose to connect a live frontier cloud key, you can obtain one from the official platforms below:

🔵 Google AI Studio (Free Tier for Gemini 2.0 Flash) → 🟢 OpenAI Platform (GPT-4o) → 🟠 Anthropic Console (Claude 3.5 Sonnet) → 🔵 DeepSeek Platform (DeepSeek V3 / R1) →

🛡️ 4 Golden Rules of Prompt Engineering

Production guidelines to eliminate hallucination and maximize signal.

1. Ban Conversational Filler

Never let the model waste compute on "Sure, I'd be happy to help!"

Eliminate conversational filler, preambles, and pleasantries. Begin directly with the deliverable.
2. Negative Boundaries

Stating what NOT to do reduces hallucinations by up to 80%.

Do not fabricate statistics or quote unverified data. If uncertain, state the boundary explicitly.
3. Enforce the XYZ Formula

Google's battle-tested standard for quantifiable accomplishments.

Every accomplishment bullet MUST follow: Accomplished [X], measured by [Y], by doing [Z].
4. Enforce Explicit Headers

Declaring structure yields reproducible, professional formatting.

Format strictly as Markdown: 1. Executive BLUF, 2. Core Trade-Offs, 3. Decision Matrix.

Ready to Build High-Performance Prompts?

Put the RTC-CO framework and the 12 domain engines to work in our free, privacy-first prompt workspace.

⚡ Open Draft a Prompt Workspace →