# Productization Plan: Research-to-Deck Skill Suite

## Vision
A suite of skills that takes a project brief and produces a complete, presentation-ready deck overnight. Research, strategy, content, visuals, layout — all automated through the agent team with hard-enforced quality gates.

**Input:** "Build me a deck for [project/company/product]"
**Output:** Figma deck with researched content, strategic positioning, consistent illustrations, professional layout.

---

## What We Know How to Do (Proven Feb 3-5)

| Capability | Proven On | Current State |
|-----------|-----------|---------------|
| Visual/competitive research | PHAT Foods, Bonanzo | `visual-research` skill exists |
| Brand positioning & strategy | PHAT Foods | Manual (Ogilvy + Kitt) |
| Content planning (slide maps) | PHAT Foods (50-slide map) | Manual markdown |
| Figma template automation | PHAT Foods deck | `figma-deck-workflow.md` documented |
| Style-consistent illustration | Bonanzo (Ozawa style) | `deck-illustrations` skill exists |
| Quality critique pipeline | All projects | WORKFLOWS.md, but not enforced |
| Competitive intelligence | PHAT Foods (9 competitors) | Manual research |

---

## Architecture: The Skill Suite

```
┌─────────────────────────────────────────────┐
│           DECK FACTORY (orchestrator)         │
│  Takes a brief → runs the full pipeline      │
│  Stakeholder: Kitt                            │
└──────────┬──────────────────────────────────┘
           │
    ┌──────┴──────┐
    │             │
┌───▼───┐   ┌────▼────┐
│RESEARCH│   │STRATEGY │
│ PHASE  │   │  PHASE  │
└───┬────┘   └────┬────┘
    │             │
    └──────┬──────┘
           │
    ┌──────▼──────┐
    │  PRODUCTION  │
    │    PHASE     │
    └──────┬──────┘
           │
    ┌──────▼──────┐
    │     QA       │
    │    PHASE     │
    └─────────────┘
```

---

## The Skills

### Layer 0: Infrastructure (Fix Criticals)

#### Skill: `pipeline-enforcer`
**What:** Hard enforcement of delegation. Kitt's only interface to creative/production work.
**Stakeholder:** Kitt
**Rules:**
- Kitt CANNOT run image generation, write copy, or build slides directly
- All creative tasks go through this skill, which spawns the right agent
- Blocks completion until Anton-approved output exists
- Logs every pipeline execution for audit

#### Skill: `flat-orchestrator`
**What:** Pattern for multi-agent work without nested spawning.
**Stakeholder:** Kitt
**Rules:**
- Kitt spawns all agents directly (no agent-spawns-agent)
- Defined handoff protocol: output of Agent A becomes input to Agent B
- Built-in retry logic (5x before escalating)
- Timeout handling with graceful fallback

#### Skill: `output-verifier`
**What:** Mandatory visual verification before any deliverable is reported complete.
**Stakeholder:** Anton
**Rules:**
- Takes a screenshot/snapshot of every visual output
- Runs checklist: readable? correct colors? proper layout? matches brief?
- Blocks "done" status until verification passes
- Generates before/after comparison when iterating

#### Skill: `pre-flight`
**What:** Validates all dependencies before production.
**Stakeholder:** Kitt
**Rules:**
- Checks: API keys valid, services reachable, plugin connected, channel confirmed
- Runs automatically before any production workflow
- Fails fast with clear error messages
- Can be run standalone for diagnostics

### Layer 1: Research Phase

#### Skill: `market-research` (new)
**What:** Comprehensive competitive & market research for any industry/product.
**Stakeholder:** Julia (research agent)
**Input:** Product/company name, industry, known competitors
**Output:** Structured research doc with:
- Market size & growth
- Key players (with 1-liner positioning for each)
- Industry trends (3-5)
- Gaps/opportunities
- Consumer insights
- Sources for every claim
**Rules:**
- Every fact must have a source
- Distinguish between hard data and estimates
- Flag where data is thin or conflicting
- Output format: structured markdown with citation links

#### Skill: `visual-research` (exists — improve)
**What:** Competitive visual/brand analysis.
**Stakeholder:** Tatiana
**Input:** Industry/product, competitor list, brand direction
**Output:** Visual research doc with:
- Competitor brand analysis (logo, palette, typography, imagery style)
- Design trends in the space
- White space / opportunities for differentiation
- Mood board brief (what to emulate, what to avoid)
**Rules:**
- Screenshots/links for every reference
- Strategic reasoning for each observation (not just "looks nice")
- Explicit recommendations, not just observations

### Layer 2: Strategy Phase

#### Skill: `brand-strategist` (new)
**What:** Takes research output and produces brand positioning.
**Stakeholder:** Ogilvy (copy) + Kitt (strategy)
**Input:** Market research doc, visual research doc, client brief
**Output:**
- Positioning statement (10-word, manifesto, elevator pitch)
- Brand pillars (3-5, with proof points)
- Competitive killers (1-liner per competitor)
- Tone of voice guide
- Naming directions (if needed)
**Rules:**
- Every claim must be backed by research
- Competitive killers must be specific and brutal (not generic)
- Tone guide must include examples + anti-examples
- Output feeds directly into content-planner

#### Skill: `content-planner` (new)
**What:** Takes strategy output and produces a complete slide map.
**Stakeholder:** Kitt + Ogilvy
**Input:** Brand strategy doc, target audience, presentation context (investor, client, internal)
**Output:** Content bank markdown with:
- Slide-by-slide plan (number, title, template type, content, image brief)
- Narrative arc annotation (setup → tension → resolution)
- Speaker notes
- Source attributions per slide
**Rules:**
- Every slide has a purpose (no filler)
- Template assignment from available template registry
- Image briefs are specific enough for Tatiana to execute without questions
- Max 20-25 content slides (plus cover, dividers, closing)

### Layer 3: Production Phase

#### Skill: `deck-builder` (exists — rebuild)
**What:** Takes content bank + templates and builds the Figma deck.
**Stakeholder:** Kitt (automation)
**Input:** Content bank markdown, Figma template registry, style brief
**Output:** Complete Figma deck with all slides built as component instances
**Rules:**
- Uses template registry (component IDs, not clone)
- Layout follows convention (rows for chapters, columns for slides)
- All text populated, all images placed
- Pre-flight check before starting (plugin connected, templates exist)
- Batch execution (not one-at-a-time)

#### Skill: `illustration-pipeline` (rebuild from deck-illustrations)
**What:** Generates all deck illustrations in a consistent style.
**Stakeholder:** Tatiana + Anton
**Input:** Image briefs from content bank, style lock doc
**Output:** Generated images, placed in Figma
**Rules:**
- Style lock doc must exist before ANY generation starts
- Character sheet established on first image, enforced on all subsequent
- Anton reviews every image before placement
- Batch processing: generate all, review all, place all
- Maintain approved asset library per project

#### Skill: `style-lock` (new)
**What:** Creates and maintains style consistency documents.
**Stakeholder:** Tatiana
**Input:** Reference images, brand direction, art style name
**Output:** Style lock document with:
- Character sheet (proportions, clothing, accessories, colors)
- Background treatment (texture, color palette, lighting)
- Camera angles and composition rules
- Do/don't list with examples
- Prompt template (fill-in-the-blank for each new image)
**Rules:**
- Created BEFORE any image generation
- Updated when new elements are approved
- Referenced in every image generation prompt
- Anton validates style lock before production starts

### Layer 4: QA Phase

#### Skill: `deck-qa` (new)
**What:** Full quality review of completed deck.
**Stakeholder:** Anton
**Input:** Completed Figma deck
**Output:** QA report with:
- Slide-by-slide check (text readable, layout correct, images present)
- Brand consistency check (colors, fonts, tone)
- Content accuracy check (facts, sources, numbers)
- Visual consistency check (style lock compliance)
- List of fixes needed (with priority)
**Rules:**
- Takes actual screenshots, not just metadata
- Compares against content bank (was everything included?)
- Compares against style lock (is visual language consistent?)
- Blocks delivery until all critical issues resolved

---

## Stakeholder Map

| Agent | Role in Factory | Skills They Own |
|-------|----------------|-----------------|
| **Kitt** | Orchestrator, strategic framing, final review | pipeline-enforcer, flat-orchestrator, pre-flight, deck-factory |
| **Julia** | Research | market-research |
| **Ogilvy** | Copy, positioning, content planning | brand-strategist, content-planner |
| **Tatiana** | Visual execution, style management | visual-research, illustration-pipeline, style-lock |
| **Anton** | Quality gates at every phase | output-verifier, deck-qa |
| **Erica** | Brief writing, planning docs | Assists on content-planner, coordinates handoffs |

---

## Execution Plan

### Phase 1: Fix Criticals (Day 1)
1. Build `pipeline-enforcer` skill — hard block on Kitt doing creative work
2. Build `flat-orchestrator` skill — solve the sub-agent nesting problem
3. Build `output-verifier` skill — mandatory visual verification
4. Build `pre-flight` skill — dependency validation
5. Test: Run a simple creative task through the enforced pipeline

### Phase 2: Research Skills (Day 2)
1. Build `market-research` skill — structured competitive research
2. Improve `visual-research` skill — add strategic reasoning, output format
3. Build `style-lock` skill — style consistency documents
4. Test: Run full research phase for a test project

### Phase 3: Strategy Skills (Day 3)
1. Build `brand-strategist` skill — positioning from research
2. Build `content-planner` skill — slide maps from strategy
3. Test: Research → Strategy pipeline end-to-end

### Phase 4: Production Skills (Day 4)
1. Rebuild `deck-builder` skill — template-driven Figma automation
2. Rebuild `illustration-pipeline` skill — style-locked image generation
3. Build `deck-qa` skill — full quality review
4. Test: Content bank → Built deck → QA'd output

### Phase 5: Integration (Day 5)
1. Build `deck-factory` orchestrator — chains all phases
2. End-to-end test: Brief → Research → Strategy → Content → Production → QA → Delivered deck
3. Document the full system
4. Stress test with a new project (not PHAT Foods or Bonanzo)

---

## Success Criteria

**The factory works when:**
- Assaf gives a 1-paragraph brief
- Overnight (or in a few hours), the system produces:
  - Researched, sourced competitive analysis
  - Sharp positioning and brand strategy
  - Complete content bank with slide map
  - Style-consistent illustrations
  - Built Figma deck with all content populated
  - QA report with any remaining issues flagged
- Assaf reviews ONE deliverable (the deck), not 15 intermediate artifacts
- Quality hits the bar on first delivery >80% of the time