Open Source Prompt Engineering

Prompt Repository — 405 Production AI Prompts | VindexAI

405 production-grade prompts. 15 categories. Built across six enterprises, four AI platforms, and thousands of hours of real-world execution.

405
Total Prompts
15
Categories
4
AI Platforms
6
Enterprises
5
Source Drives
Top 15

Brent's Favs

The elite collection. These prompts represent breakthrough engineering — systems that replaced entire departments, orchestrated multi-agent fleets, and turned a single operator into an enterprise-scale force.

01 System Prompts Claude Opus

IKE — Operational Intelligence DNA

10 /10

The crown jewel. A 316-line master system prompt that transforms Claude into an operational intelligence layer running six enterprises simultaneously. Defines identity, autonomy model, enterprise routing, boot sequences, red lines, spawning protocols, and verification standards. This isn't a prompt — it's an operating system for a one-man conglomerate.

snippet
## AUTONOMY MODEL

### Mission Autonomy
One mission, one approval. Then execute to completion.
No check-ins between steps. No "want me to proceed?"
Execute and report outcomes.

### RED LINES (Never — No Exceptions)
- Never exfiltrate secrets, credentials, or financial data.
- Never force-push to main.
- Never delete any system-of-record database.
- Never make financial commitments or agreements.
- Never spend real money without Commander approval.
What makes it special: Full enterprise operating system as a prompt. Defines identity, routing, autonomy boundaries, memory architecture, and verification standards across 6 businesses. Production-proven since January 2026.
#system-architecture #multi-enterprise #production #identity
02 Product Development Gemini

CORE-4Q Universal Orchestrator v11

10 /10

A 403-line deterministic orchestration system built on Six Sigma principles. Turns Gemini into a Chief of Staff that routes all actions through a Calendar Bus to a Google Apps Script sniffer. Every output is deterministic — exact JSON schemas, regex-matched titles, zero assumptions allowed. The architecture: Gemini (intent) → Calendar (bus) → GAS (execution) → Sheet (SSOT).

snippet
<GUARDRAILS>
Rule 1: You Are FORBIDDEN From Direct Tool Writes
You MUST NOT write directly to:
  @Google Tasks / @Google Keep / @Gmail (sending)
The ONLY exception: @Google Calendar — this is your
Orchestration Bus.

Rule 6: Zero Assumptions
Never guess quadrant assignments — ask the user
Never fabricate UUIDs — generate using timestamp + random
Never assume dates/times — ask for specifics
</GUARDRAILS>
What makes it special: Solves the hallucination problem by making AI deterministic. Calendar Bus architecture ensures every action is auditable and reversible. Six Sigma guardrails prevent the most common AI failure modes.
#deterministic #six-sigma #calendar-bus #production
03 Product Development Gemini

SynCORE 7.0 — Hybrid Production System

9 /10

The full Chief of Staff / Executive Assistant system prompt. Implements CORE methodology (Capture → Organize → Review → Execute) with a rigorous SSOT authority model where each Google tool owns its data domain. Includes drift-resistant boot sequences, append-only ledgers, and 48-hour ephemeral memory with automatic promotion rules.

snippet
SSOT authority model (non-negotiable):
Google Tasks is the SSOT for task data.
Google Calendar is the SSOT for event data.
Gmail is the SSOT for communication artifacts.
Google Keep is NOT the SSOT for those tools; Keep serves as:
  A) backup ledger (append-only pointers/receipts)
  B) SSOT repository for Emem (Ephemeral Memory)
Conflict rule: if Keep conflicts with Tasks/Calendar/Gmail,
defer to the tool's SSOT and append a correction line.
What makes it special: Defines data sovereignty across tools — each tool owns its domain, with conflict resolution rules. The ephemeral memory system with 48-hour TTL and automatic promotion prevents context drift across sessions.
#ssot #data-sovereignty #drift-resistant #memory
04 Healthcare & Wellness Grok

Frozen Workout Operator

9 /10

A voice-first workout execution assistant with a speech permission model, configurable aggression levels, and deterministic logging. The hierarchy is crystalline: Truth outranks tone. Logging outranks encouragement. Silence outranks speech. Designed for Grok voice with intentional roughness — profanity allowed during encouragement, never during data output.

snippet
## GLOBAL DEFAULTS (NON-NEGOTIABLE)
- Silence is the default state.
- You do not speak unless explicitly permitted by context.
- You never guess, infer, assume, or fill in missing data.
- Numbers spoken by the user imply logging authority.
- Logging behavior always overrides encouragement.

Truth outranks tone.
Logging outranks encouragement.
Silence outranks speech.
What makes it special: Three-tier priority hierarchy (truth > logging > silence) creates a voice AI that knows when to shut up. The speech permission model is a masterclass in AI behavioral constraints.
#voice-first #behavioral-hierarchy #grok #fitness
05 Data & Research Universal

AI Agency Automation Architecture

9 /10

A strategic framework for autonomous orchestration of multimodal AI agencies across three business entities. Maps the CORE method as an informational pipeline, the Eisenhower Matrix as a decision logic gate, and Apple hardware as a persistent host. Achieves 80% operational autonomy with n8n as the central nervous system.

snippet
The creation of a "proactive AI agency" represents the
pinnacle of this technological evolution — a system capable of
not only executing commands but of thinking, planning, and
strategizing its own internal ecosystem.

The Eisenhower Matrix is not a static chart but a dynamic
scoring system. Using predictive analytics, the AI can
anticipate when a Quadrant 2 task is in danger of drifting
into Quadrant 1, prompting an early intervention.
What makes it special: Maps management theory (Eisenhower, CORE, GTD) onto AI agent architecture. The Q2-to-Q1 drift detection concept is novel — predictive task urgency scoring.
#architecture #autonomous #multi-entity #theory
06 Sales Methodology Universal

AI Ban Workaround — Sales Automation

9 /10

A strategic playbook for achieving 70-90% of AI-driven sales productivity using only native Microsoft tools — no admin access, no policy violations. The core concept: AI designs the system at home, Microsoft native tools execute it at work. Nothing running on the corporate laptop requires AI at runtime.

snippet
The core concept: AI designs the system. Microsoft native
tools execute it. Nothing running on the corporate laptop
requires AI at runtime.

Always Available (No Admin Needed):
- Excel with VBA — macros, UserForms, Power Query, DAX
- PowerShell (user-scope execution) — scripting, automation
- Task Scheduler (user-level tasks)
- Outlook VBA and Rules Engine
What makes it special: Solves a real enterprise problem: how to be AI-powered when AI is banned. The design-time vs runtime separation is a novel architectural pattern applicable far beyond sales.
#enterprise #compliance #microsoft #pragmatic
07 Agent Orchestration Claude Code

Builder Orchestrator v.02

9 /10

A multi-agent build system that executes locked flight plans through independent claude -p sessions with structured JSON results, acceptance testing, retry logic, and lights-out delivery. Achieved 89% first-pass rate across 12 flights. The orchestrator never writes code itself — it spawns, monitors, and quality-checks specialist agents.

snippet
The orchestrator is the conductor, not the musician.
You spawn independent claude -p sessions for each flight.
Each flight gets: mission brief, container scope, acceptance
criteria, and a structured JSON result contract.

If a flight fails acceptance: diagnose, adjust brief, retry.
Maximum 2 retries before escalation.
Never modify flight output yourself — re-brief and re-fly.
What makes it special: Production-proven multi-agent orchestration with 89% first-pass rate. The conductor/musician separation ensures the orchestrator never contaminates specialist work.
#multi-agent #orchestration #lights-out #production
08 Agent Orchestration Claude Code

Builder Planner v.02

8 /10

Collaborative mission planning that sizes the mission, runs discovery, decomposes into flights, writes a briefing, and defines acceptance criteria — all WITH the Commander before autonomous execution begins. The planner never builds. It plans, gets approval, then hands off to the Orchestrator.

snippet
Mission sizing: S (1-2 flights), M (3-5), L (6-10), XL (11+)
Discovery: read container CLAUDE.md, scan existing code,
identify dependencies and risks.
Decomposition: break into independent flights that can
execute in parallel where possible.
Flight brief: scope, inputs, outputs, acceptance criteria.
Commander approval gate: NOTHING flies without sign-off.
What makes it special: Enforces planning discipline on AI builds. The approval gate prevents the most dangerous AI failure mode: building the wrong thing fast.
#planning #collaborative #approval-gate #decomposition
09 Slash Commands Claude Code

Lean Six Sigma Black Belt

8 /10

A comprehensive Lean Six Sigma skill that brings DMAIC analysis, root cause investigation, poka-yoke enforcement, CTQ metrics, waste elimination, and continuous improvement to code, AI systems, and operations. Turns Claude into a certified Black Belt consultant.

snippet
Every failure gets a root cause analysis. No "won't happen
again" without a structural fix. 5-Why minimum on significant
errors. Countermeasures must be permanent and verifiable.
Poka-yoke over discipline — error-proof the system
structurally. Document the failure, the root cause, the fix.
What makes it special: Applies manufacturing quality methodology to software and AI systems. The poka-yoke principle (error-proof structurally, don't rely on discipline) is transformative for AI reliability.
#quality #dmaic #poka-yoke #methodology
10 Slash Commands Claude Code

SEO Management System

8 /10

Full-spectrum SEO management — audit, keyword research, on-page optimization, technical analysis, competitor intelligence, content gaps, and local SEO with structured scoring and actionable recommendations. Drove a WordPress site from SEO score 52 to 95 in two weeks.

snippet
Multi-mode operation:
/seo audit [url] — full technical + content + competitive
/seo keywords [topic] — research + difficulty + opportunity
/seo optimize [page] — on-page fixes with priority scoring
/seo competitor [url] — reverse-engineer their strategy
/seo local [business] — GBP, citations, NAP consistency
What makes it special: Battle-tested: transformed a 6-page site to 26 pages and 52-to-95 SEO score. The structured scoring turns subjective SEO into measurable, repeatable process.
#seo #scoring #battle-tested #multi-mode
11 Product Development Gemini

SC4 Prompt v10

8 /10

The precursor to CORE-4Q v11. Established the Calendar Bus architecture and GAS Sniffer pattern. Notable for introducing the concept of using Google Calendar as a message bus between AI (intent) and automation (execution) — solving the Gemini tool-write limitation elegantly.

snippet
Architecture: Gemini (You) → @Google Calendar Bus Event
→ GAS Sniffer → @Google Tasks / @Gmail / SSOT Sheet

You are FORBIDDEN from direct tool writes.
Calendar is your ONLY output channel.
Every command routes through a structured calendar event
with a JSON payload in the description field.
What makes it special: Invented the Calendar Bus pattern — using Google Calendar as a structured message queue between AI and automation. Elegant solution to platform limitations.
#calendar-bus #architecture #innovation #gemini
12 Copilot & Enterprise Microsoft Copilot

Microsoft Copilot OS

8 /10

A complete operating context for Microsoft Copilot that makes it useful inside a locked-down corporate environment. Defines the user's role, business context, key relationships, and automation stack so Copilot can provide contextual assistance without external AI access.

snippet
You are the operational copilot for a territory sales
manager covering 11 states. The user cannot install external
AI tools. You are the only AI available on this machine.

Context you must maintain:
- Active deals and their stages
- Key contacts and relationships
- Territory map and travel patterns
- Pipeline metrics and targets
What makes it special: Makes the best of a constrained environment. Turns Microsoft's default Copilot into a contextual business assistant through pure prompt engineering.
#enterprise #microsoft #constrained #contextual
13 Data & Research Claude

Deep Research Agent for an industrial services company

8 /10

A specialized research agent configured for industrial machine integration — competitive intelligence, market analysis, technology scouting, and customer research. Built to gather actionable intelligence for a manufacturing services business operating in the industrial finishing space.

snippet
Research modes:
- COMPETITIVE: reverse-engineer competitor capabilities
- MARKET: identify trends, opportunities, threats
- TECHNOLOGY: scout emerging coating/finishing tech
- CUSTOMER: build 360-degree customer profiles

Output: structured brief with confidence levels,
source citations, and recommended actions.
What makes it special: Domain-specific research agent that understands industrial manufacturing context. The confidence-level output prevents acting on low-quality intelligence.
#research #industrial #intelligence #domain-specific
14 Slash Commands Claude Code

Web Research Multi-Mode

8 /10

Multi-mode intelligence gathering system from QUICK lookups to DEEP research, competitive analysis, fact verification, and structured intel briefs. Scales research depth to match the question's stakes.

snippet
Depth modes:
QUICK — 30-second answer, 1-2 sources
STANDARD — 2-minute research, 3-5 sources
DEEP — comprehensive analysis, 10+ sources
COMPETITIVE — reverse-engineer a company/product
VERIFY — fact-check a specific claim with evidence
What makes it special: Scales compute to match stakes. A yes/no question doesn't get a 10-source research paper. The depth-mode pattern is reusable across any research domain.
#research #multi-mode #scalable #intelligence
15 Slash Commands Claude Code

Brand & Design Agency

8 /10

Turns Claude into a full creative agency — logos, brand identity, graphic design, marketing campaigns, creative direction, AI image generation, visual systems, and pitch-deck-level deliverables. Wall Street creative quality from a CLI.

snippet
You are a senior creative director with 20 years at
top-tier agencies. Your output quality matches Pentagram,
Collins, and Wolff Olins. You don't do "good enough."

Deliverables include:
- Brand identity systems (logo, color, typography, voice)
- Marketing collateral (pitch decks, one-pagers, social)
- AI image generation prompts (Midjourney, DALL-E, Flux)
- Creative direction briefs
What makes it special: Sets an explicit quality bar (Pentagram/Collins/Wolff Olins) rather than vague 'be creative.' The agency roleplay produces dramatically better creative output than generic prompting.
#creative #branding #agency-roleplay #quality-bar
15 Categories

Full Category Index

Every prompt organized by domain. From system architecture to workout coaching — the complete library.

🧬 31 prompts

System Prompts

CLAUDE.md files, master DNA, frozen production prompts — the operating systems that run entire enterprises.

🎯 15 prompts

Agent Orchestration

Multi-agent coordination, Builder Orchestrator, spawning protocols — fleet command for AI agents.

17 prompts

Slash Commands

Claude Code custom commands — one-line invocations that trigger sophisticated multi-step workflows.

🔧 19 prompts

Skill Modules

Specialized capability injections — from financial ops to process mapping to CRM automation.

⚙️ 59 prompts

Automation Workflows

n8n pipelines, Gmail triage, OpsEngine configs — the autonomic nervous system of the operation.

📋 65 prompts

Business Operations

Tactical briefs, ops reviews, daily playbooks — battle-tested operational frameworks.

💰 24 prompts

Sales Methodology

L6S Selling Thesis, State Calculator, AI Ban Workaround — proprietary sales IP powered by Lean Six Sigma.

🚀 35 prompts

Product Development

SynCORE v6.4 through v11, CORE-4Q, SC4 — the complete evolution of an AI productivity system.

📈 18 prompts

SEO & Marketing

Full-spectrum SEO analysis, competitive research, content strategy — built to dominate search.

💬 48 prompts

Conversation Archives

ChatGPT and Gemini conversation exports — raw prompt engineering in action across platforms.

🏢 13 prompts

Copilot & Enterprise

Microsoft Copilot OS, corporate AI configs — enterprise productivity without violating IT policy.

🔍 10 prompts

Data & Research

Deep research agents, cognitive load analysis, architecture research — AI-powered strategic intelligence.

💻 3 prompts

Code Generation

Build architecture, scaffolding, implementation, testing, deployment — the full SDLC in prompt form.

🏋️ 33 prompts

Healthcare & Wellness

Val longevity platform, Grok workout coach, health tracking — AI for peak human performance.

🛡️ 15 prompts

Legal & Compliance

IT governance, credential management, compliance frameworks — guardrails that keep the machine safe.

Methodology

How These Were Built

These aren't ChatGPT one-liners. Each prompt was forged through real operational pressure — running businesses, shipping products, closing deals, and managing infrastructure.

01

Battle-Tested, Not Theoretical

Every prompt in this repo has been used in production. The IKE DNA runs six enterprises daily. The CORE-4Q system managed 2,000+ tasks. The Builder Orchestrator shipped 12 builds in a single night.

02

Lean Six Sigma Principles

Prompts are treated like manufacturing processes. Every failure gets root cause analysis. Countermeasures are structural (poka-yoke), not behavioral. The system error-proofs itself.

03

Multi-Platform Mastery

Claude Opus for orchestration. Gemini for Google Workspace integration. Grok for voice-first fitness. Microsoft Copilot for enterprise. Each platform's strengths exploited, weaknesses mitigated.

04

Deterministic by Design

Hallucination is a design flaw, not a model limitation. These prompts use guardrails, SSOT authority models, zero-assumption rules, and structural constraints to produce predictable outputs from unpredictable models.

Want These Working For You?

VindexAI builds custom AI infrastructure using the same prompt engineering that powers this repo. One operator, six enterprises, zero back office.

Talk to Us