# M-06 — Alignment / shared mental model
# Category: MUST
# Kapitel: 10 – Alignment des Verständnisses
# Anker: korrekt ausgeführtes Missverständnis; Teach-back
# Orientierung: 16:9 landscape
# Didaktik: Mensch und Agent brauchen Shared Model vor großen Diffs; 5-Minuten-Protokoll.
# Caption (DE): Shared Model vor Execute — sonst bleibt der teuerste Fehler ein korrekt ausgeführtes Missverständnis.
# Ziel-Asset: assets/illustrations/M-06-alignment-shared-model.jpg
# Language: English (Gemini / Imagen)
# Usage: paste entire prompt body below into Gemini

--- PROMPT START ---
Create a conceptual alignment diagram, 16:9 landscape, for a professional practice book about coding agents.

TOPIC: The costliest agent failure is correctly executing a misunderstanding. Build and keep a shared model before big diffs.

GLOBAL STYLE:
- editorial infographic; navy/teal Venn; calm off-white panels; flat vector
- minimal English labels; print-ready; abstract geometric avatars only if needed (no photoreal faces)

QUALITY GUARDRAILS:
- every visible word correctly spelled English; max ~12 short labels
- no brand logos; no cute robots

Title text exactly: "Alignment of understanding"
Subtitle: "correctly executed misunderstanding"

CENTER: Venn of two circles "Human" and "Agent"
Intersection bright teal: "Shared model"
chips inside intersection only: "outcome · scope · invariants · DoD"
Dim fringes risk labels: "plausible assumptions" · "spec mutation in chat"

BELOW: 5-step protocol strip:
1 "task prompt"
2 "explore + assumptions + plan"
3 "correct / approve"
4 "execute to DoD"
5 "verify + report"
Note chip: "bad teach-back → do not execute"

BOTTOM three symptom→fix cards:
"neighbor problem diff → rescope"
"green but product says no → human gates"
"two sessions disagree → one spec file"

MOOD: calm collaboration diagram for teams.
ASPECT: 16:9.
NEGATIVE: no photoreal faces, no neon brains, no watermarks, no brand logos.
--- PROMPT END ---

# Follow-up edit:
# Keep Venn shared model + 5-step protocol + teach-back stop rule.
