---
# NAME/DIRECTORY (Phase 234-04, Agent Skills spec): this skill shipped as
# `skills/MOSDeckEngine/` with `name: MOSDeckEngine` until 2026-07-28. The Agent
# Skills spec (agentskills.io/specification) requires `name` to match
# /^[a-z0-9]+(-[a-z0-9]+)*$/ AND to equal the parent directory, so a CamelCase
# name made this skill vanish from a strict foreign host's catalog. Phase 234-04
# renamed the directory to `mos-deck-engine` and the `name:` field with it. The
# LEGACY ALIAS HANDLE `MOSDeckEngine` in data/deck-aliases.json is deliberately
# UNCHANGED: that key is a back-compat routing handle a navigator may still type,
# not a filesystem identifier, and retiring it would break the Phase 175
# deprecate-not-delete contract.
name: mos-deck-engine
description: >
  Use ONLY when the navigator explicitly asks to build a pitch deck, investor
  presentation, or fundraising deck (or names this pipeline directly). Do NOT use
  for a casual "explain this more simply" or "make this clearer" remark with no
  deck or presentation context -- that stays in conversation. Runs a 6-stage
  Feynman first-principles pipeline: reduce to essence, translate to plain
  language, expose confusion, build mental models, simplify until it breaks,
  teach it back. Outputs YC-quality slides. Triggers: pitch deck, investor
  presentation, simplify for investors, technical storytelling, demo day,
  fundraising deck.
license: BSL-1.1. See LICENSE for complete terms (Business Source License 1.1, Change Date 2030-04-16 to Apache License 2.0).
# --- Phase 143.3 connector frontmatter ---
connector:
  connects_to_spine: true
  sensor_triggers: []
  reach_id: context_block
  sub_mode: deck-engine
  posture: hold
  hierarchy_rank: 14
  filing: memory_event_only
  plan_gated: false
  web_scope: null
  surface: F.1
hitl_shape: "F.1"
hitl_why: "Deprecated redirect to /mos:deck; inherits /mos:deck's F.1 Next Move gate at the close of the Feynman pipeline stages, never introduces a separate fork of its own."
layer: "loop"
layer_why: "Runs a single 6-stage Feynman simplification cycle end to end and files a deck artifact; one continuous methodology run, not a multi-agent or multi-gate coordination."
---

# mos-deck-engine -- DEPRECATED, consolidated into /mos:deck

> **Deprecation redirect (Phase 175, deprecate-not-delete).**
> This engine is now consolidated into the `/mos:deck` command. Invoke `/mos:deck`
> and pick the **Feynman** style to run the 6-stage first-principles comprehension
> pipeline this skill used to run on its own. The `/mos:deck` command also adds the
> HEART persuasion spine and a mesh composite, a methodological per-section build
> flow, and a WARN-first deck-design ruleset (source links, brand auto-binding with
> the logo linking to https://mindrian-os.com, AI-image provenance footers).
>
> This SKILL.md is RETAINED for back-compat (the `MOSDeckEngine` handle still
> triggers and still routes to `/mos:deck` via `data/deck-aliases.json`), and the
> `connector:` block above is preserved so the surface stays WIRED under CIRS. New
> work should invoke `/mos:deck` and pick the Feynman style. The original 6-stage
> Feynman pipeline (reduce to essence, translate, expose confusion, build mental
> models, simplify until it breaks, teach it back) now lives behind the Feynman
> style of `/mos:deck`; the canonical stage list is `data/deck-styles.json`
> (`feynman_stages[]`).

## Where the engine went

| Was | Now |
|-----|-----|
| `mos-deck-engine` skill (this file, formerly `MOSDeckEngine`) | `/mos:deck` command, Feynman style |
| `feynman-engine` skill (out-of-repo) | `/mos:deck` command, Feynman style |
| Standalone 6-stage Feynman pipeline | Feynman style spine inside `/mos:deck` |
| Pitch / investor / demo-day deck | `/mos:deck` (Feynman or HEART or mesh) |

Both prior handles (`MOSDeckEngine`, `feynman-engine`) resolve to `/mos:deck` per
`data/deck-aliases.json` (D-04b). Pick the Feynman style for the comprehension
spine this skill was built for.

---

## Appendix: the original 6-stage Feynman pipeline (now the /mos:deck Feynman style)

The stages below are retained verbatim as reference for what the Feynman style of
`/mos:deck` runs. They are no longer invoked from this file; invoke `/mos:deck` and
pick the Feynman style.

## Stage 1: REDUCE TO ESSENCE

**Goal:** Strip the concept to irreducible fundamental truths.

**Run this prompt against the input:**

> Remove all jargon, assumptions, and surface-level explanations from this concept.
> Break it into its most fundamental truths -- pieces that cannot be simplified
> further without losing meaning. Show how they connect logically. Identify what
> is ESSENTIAL vs what is DECORATIVE.

**Produce:**
- Bullet list of irreducible truths (max 5-7)
- Connection map showing how pieces relate
- One-paragraph essence summary (3-4 sentences max)

**Challenge the user:** "That's not a fundamental truth -- that's an implementation
detail. What's underneath it?"

**Gate:** Ask "Does this capture the real core? What did I miss?" -- wait for response.

---

## Stage 2: TRANSLATE TO PLAIN LANGUAGE

**Goal:** Convert the essence into language a non-technical investor absorbs on first read.

**Run this prompt against Stage 1 output:**

> Rewrite using simple words, short sentences, everyday language. Replace abstract
> terms with concrete descriptions. Target: a smart generalist who sees 1,000
> pitches a year. No academic tone. Conversational clarity that feels obvious
> on first read.

**Produce:**
- The "elevator version" (30 seconds, 3-4 sentences)
- Glossary of unavoidable terms, rewritten simply
- The "12-year-old version" (stripped to absolute minimum, still true)

**Challenge:** "If you need the word 'leverages' to explain this, you don't
understand it yet."

**Gate:** Ask "Read the 12-year-old version. Does it still feel true?" -- wait for response.

---

## Stage 3: EXPOSE CONFUSION

**Goal:** Ruthlessly find hidden weakness in the simplified version.

**Run this prompt against Stage 2 output:**

> Examine every sentence for vagueness, skipped steps, or buried assumptions.
> Pinpoint exactly where the explanation breaks and WHY. Turn every weak spot
> into a specific question that forces deeper thinking. Do NOT accept partial
> understanding.

**Produce:**
- Numbered gap list with severity (CRITICAL / MODERATE / MINOR)
- Root cause of each gap
- Priority questions that must be answered
- Investor impact: which gaps lose a YC partner in 3 seconds?

**Challenge:** "Paragraph two -- you jumped from problem to solution without
explaining why THIS solution. An investor notices that in 3 seconds."

**Gate:** This stage LOOPS. Fix critical gaps, re-expose, repeat until clean.
Expect 2-3 iterations. Only proceed when no CRITICAL gaps remain.

---

## Stage 4: BUILD MENTAL MODELS

**Goal:** Create analogies that make the concept instantly graspable.

**Run this prompt against the gap-fixed output:**

> Design 2-3 powerful analogies that map this concept to familiar experiences.
> Each model must preserve the STRUCTURE of the original idea while making it
> intuitive. Map each analogy element to the real element. Identify where each
> analogy breaks down.

**Produce:**
- 2-3 mental models with explicit mapping (analogy -> reality)
- Limitations of each (where it stops being accurate)
- Recommendation: best for 3-minute vs 30-minute explanation

**Challenge:** "The best analogy isn't clever -- it's obvious. When someone
hears it, they should think 'of course.'"

**Gate:** Ask "Which analogy resonates most? We'll lead the deck with that one."
-- wait for response.

---

## Stage 5: SIMPLIFY UNTIL IT BREAKS

**Goal:** Find the exact boundary between useful simplification and distortion.

**Run this prompt against the chosen model:**

> Take the best explanation and simplify it further, step by step. At each level,
> check: is the meaning still preserved? Push until further reduction WOULD
> distort the truth. Identify the exact boundary and explain why it exists.

**Produce:**
- Simplification ladder (4-5 rungs from detailed to ultra-simple)
- The breaking point: what cannot be simplified further and why
- The sweet spot: simplest version that is still TRUE

**Challenge:** "We just crossed the line. That last simplification turned your
distributed ledger into 'a shared spreadsheet' -- and that's wrong in a way
that matters. Step back one."

**Gate:** Ask "This is your presentation language. Sound right?" -- wait for response.

---

## Stage 6: TEACH IT BACK (Quality Gate)

**Goal:** Validate the final version stands alone without external context.

**Run this prompt against the sweet-spot version:**

> Present this concept from scratch. Start from basics, build up logically.
> Target: smart generalists. Evaluate for clarity, completeness, logical flow,
> persuasiveness. The listener should be able to explain this to someone else.

**Produce:**
- Polished teach-back version (this becomes the deck narrative)
- Remaining weakness (if any)
- Confidence rating: LOW / MEDIUM / HIGH / YC-READY

**Challenge:** "Read this out loud. If you stumble anywhere, that's where the
investor zones out."

**Final gate:** Only proceed to deck generation if confidence is HIGH or YC-READY.
If MEDIUM, loop back to the weakest stage. If LOW, restart from Stage 3.

---

## Deck Generation

After all 6 stages pass, generate a presentation deck.

### Slide Architecture (10-12 slides)

| # | Slide | Content Source |
|---|-------|---------------|
| 1 | Title | Essence summary (Stage 1) |
| 2 | The Problem | First-principles WHY this matters (Stage 1) |
| 3 | The Insight | The "aha" from Stage 5's sweet spot |
| 4 | The Solution | Plain language (Stage 2) + best analogy (Stage 4) |
| 5 | How It Works | Mental model DIAGRAM -- not technical architecture |
| 6 | Why Now | Simplified causal chain from Stage 1 fundamentals |
| 7 | Market | Numbers with Stage 2 plain language context |
| 8 | Business Model | One sentence + one visual |
| 9 | Traction | Evidence that validates the simplified story |
| 10 | Team | Why THIS team for THIS simplified problem |
| 11 | The Ask | What money buys, in sweet-spot language |

### Design Principles

- **One idea per slide.** Two sentences = two slides.
- **Headline carries the Feynman message.** Supporting text adds one layer of detail.
- **Diagrams over bullets.** Mental models from Stage 4 become visuals.
- **Whitespace is confidence.** Crowded slides signal unclear thinking.

### Visual Specifications

Generate as responsive HTML presentation:
- 16:9 aspect ratio with mobile/tablet breakpoints
- Keyboard navigation (arrows, space, home, end)
- Clean, minimal design -- investor-grade, not conference-talk flashy

**Color palette (professional, high-contrast):**
- Background: #FAFAFA (light) or #0A0A0A (dark)
- Primary text: #1A1A1A (light) or #F5F5F5 (dark)
- Accent: one bold color that matches the venture's energy
- Data/highlight: one complementary color

**Typography:**
- Headings: bold sans-serif (Inter, Helvetica Neue, or system)
- Body: regular sans-serif, 18px+ for readability at distance
- Data: monospace for numbers and metrics

**If ui-ux-pro-max skill is available**, use it for richer design intelligence:
```bash
python3 skills/ui-ux-pro-max/scripts/search.py "investor pitch deck minimal professional" --design-system
```

### Output

Save deck as HTML file. Report location to user.

---

## Adapting to Context

### With MindrianOS Data Room
- Pull technical content from room sections automatically
- File the Feynman analysis as a room artifact
- Cross-reference simplified concepts with existing room entries
- Use De Stijl design system for deck visuals

### Without MindrianOS (Standalone)
- Accept input via conversation or pasted text
- Output deck to current working directory
- No filing, no room integration
- Still runs the full 6-stage pipeline at identical quality

### With Brain Connection (MindrianOS + Brain)
- Query Brain for cross-domain analogies (Stage 4 enrichment)
- Pull SUCCESs Framework patterns for stickiness testing
- Use Minto Pyramid from Brain for SCQA structure validation
- Surface Golden Circle framing for the "Why Now" slide

---

## Common Patterns

### Technical Architecture -> Investor Deck
Input: "We built a distributed event-sourced system with CQRS and saga orchestration"
Stage 1 essence: "We built a system where every change is recorded, decisions are separated from data storage, and complex multi-step processes complete reliably even if parts fail"
Stage 5 sweet spot: "Think of it like a flight recorder for your business -- every decision is tracked, nothing gets lost, and the system heals itself"

### Research Paper -> Demo Day Pitch
Input: 3,000-word technical paper with citations
Stage 2 elevator: 4 sentences, zero jargon
Stage 4 analogy: maps to something every YC partner has experienced
Stage 6 teach-back: the partner could explain it to their colleagues

### Complex Algorithm -> Board Presentation
Input: ML model with attention mechanisms and transformer architecture
Stage 5 breaking point: "AI that reads" (too simple -- loses the mechanism that matters)
Stage 5 sweet spot: "AI that reads every document simultaneously and finds what connects them" (simplified but structurally true)

---

## Quality Standards

A Feynman-processed deck meets ALL of these:

- [ ] No jargon survives without a plain-language companion
- [ ] Every slide passes the "say it out loud" test
- [ ] At least one mental model that a non-expert remembers a week later
- [ ] The 12-year-old version is true (not just simple)
- [ ] Stage 3 found and fixed at least 2 gaps (if it found zero, the analysis was weak)
- [ ] The sweet spot is identified -- not the simplest version, but the simplest TRUE version
- [ ] Deck tells a story, not a feature list