Prompt Packs

Prompt Packs

Overview

Build and execute multi-step AI prompt workflows. Prompt packs are collections of templated prompts that guide founders through structured analysis, generation, and validation tasks with consistent, reproducible outputs.

When to Use

• Building multi-step AI workflows
• Creating templated prompt libraries
• Implementing structured AI outputs
• Designing reusable prompt sequences
• Connecting prompts to playbooks

Prompt Pack Architecture

┌────────────────────────────────────────────────────────────────┐
│                    PROMPT PACK SYSTEM                           │
├────────────────────────────────────────────────────────────────┤
│                                                                 │
│  ┌─────────────┐    ┌─────────────┐    ┌─────────────┐        │
│  │   CONTEXT   │───▶│   ROUTER    │───▶│  PACK       │        │
│  │  Industry   │    │  Selects    │    │  Selected   │        │
│  │  Stage      │    │  Pack       │    │             │        │
│  └─────────────┘    └─────────────┘    └──────┬──────┘        │
│                                                │                │
│                                         ┌──────▼──────┐        │
│                                         │   STEPS     │        │
│                                         │ 1 → 2 → 3   │        │
│                                         └──────┬──────┘        │
│                                                │                │
│         ┌──────────────────────────────────────┤                │
│         ▼                                      ▼                │
│  ┌─────────────┐                        ┌─────────────┐        │
│  │   OUTPUT    │                        │   APPLY     │        │
│  │  Structured │                        │  to Target  │        │
│  │  JSON/Text  │                        │  (DB/UI)    │        │
│  └─────────────┘                        └─────────────┘        │
│                                                                 │
└────────────────────────────────────────────────────────────────┘

Current Prompt Packs

Total: 10 packs, 28 steps

Pack Structure

interface PromptPack {
  id: string;
  title: string;
  slug: string;
  category: 'validation' | 'pitch' | 'canvas' | 'gtm' | 'pricing' | 'ideation' | 'market';
  industry_tags: string[];
  stage_tags: string[];
  version: number;
  is_active: boolean;
}

interface PromptPackStep {
  id: string;
  pack_id: string;
  step_order: number;
  purpose: string;
  prompt_template: string;
  input_schema: JSONSchema;
  output_schema: JSONSchema;
  model_preference: 'gemini-3-flash' | 'gemini-3-pro' | 'claude-sonnet';
  max_tokens: number;
  temperature: number;
}

Prompts by Category

Pack-to-Playbook Routing

Prompt Templates

Problem Snapshot Template

You are an expert startup advisor helping a founder clarify their problem statement.

INDUSTRY: {{industry}}
COMPANY DESCRIPTION: {{company_description}}

Transform this into a structured problem statement:

1. WHO: [Specific persona with job title, company size, or demographic]
2. STRUGGLE: [The specific pain they experience today]
3. WHY NOW: [Why this problem is urgent today, not 5 years ago]

Output as JSON:
{
  "who": "...",
  "struggle": "...",
  "why_now": "...",
  "confidence": 0-100
}

Validation Scorecard Template

You are an investor evaluating a startup idea.

INDUSTRY: {{industry}}
STAGE: {{stage}}
IDEA: {{idea_text}}
PROBLEM: {{problem_statement}}

Score this idea on:
1. Problem Clarity (0-20): Is the who/struggle/why-now clear?
2. Market Size (0-15): Is this a large, growing market?
3. Solution Fit (0-15): Does the solution match the problem?
4. Competition (0-15): Is there clear differentiation?
5. Team Fit (0-10): Does this team have an edge?
6. Traction (0-15): Is there evidence of demand?
7. Business Model (0-10): Is the revenue path clear?

Output as JSON with scores, total, and top 3 risks.

Edge Function: prompt-pack

// Actions
- 'search': Find packs by category, industry, stage
- 'get': Retrieve pack with all steps
- 'list': List all active packs
- 'run_step': Execute single step with inputs
- 'run_pack': Execute all steps sequentially
- 'apply': Save outputs to target (profile, canvas, etc.)

Database Schema

CREATE TABLE prompt_packs (
  id UUID PRIMARY KEY,
  title TEXT NOT NULL,
  slug TEXT UNIQUE NOT NULL,
  category TEXT NOT NULL,
  industry_tags TEXT[],
  stage_tags TEXT[],
  version INTEGER DEFAULT 1,
  is_active BOOLEAN DEFAULT true,
  created_at TIMESTAMPTZ DEFAULT NOW()
);

CREATE TABLE prompt_pack_steps (
  id UUID PRIMARY KEY,
  pack_id UUID REFERENCES prompt_packs(id),
  step_order INTEGER NOT NULL,
  purpose TEXT NOT NULL,
  prompt_template TEXT NOT NULL,
  input_schema JSONB,
  output_schema JSONB,
  model_preference TEXT DEFAULT 'gemini-3-flash',
  max_tokens INTEGER DEFAULT 2048,
  temperature NUMERIC DEFAULT 1.0
);

CREATE TABLE prompt_runs (
  id UUID PRIMARY KEY,
  startup_id UUID REFERENCES startups(id),
  user_id UUID REFERENCES auth.users(id),
  pack_id UUID REFERENCES prompt_packs(id),
  step_id UUID REFERENCES prompt_pack_steps(id),
  inputs_json JSONB,
  outputs_json JSONB,
  model_used TEXT,
  tokens INTEGER,
  cost_usd NUMERIC,
  latency_ms INTEGER,
  status TEXT DEFAULT 'completed',
  created_at TIMESTAMPTZ DEFAULT NOW()
);

AI Model Selection

References

• PRD Section 9: Prompt Pack System
• Strategy Section 8.2: Prompt Library
• /startup-system/guides/50-mvp-prompts.md
• /startup-system/guides/53-validate-prompts.md