← Files OBI Image CreationARCHIVED FILE

skills/obi-image-project/references/create-output.md

2.78 KB · Oct 5, 2026 · 18:33 UTC

↓ Download file

# Research & Create output contract

The creator-facing deliverable must contain every field below in this order. Keep the full 1,000–2,000-word research brief in the project record; show the approximately 200-word historical context here.

## Historical research deliverables

Save the full 1,000–2,000-word factual, non-devotional research brief in the project record. Do not paste it into the default response. Use the `HISTORICAL CONTEXT` field below as the approximately 200-word shareable OBI summary, covering biblical context, culture, geography, daily life, clothing, architecture, material culture, significant debates, and clearly labeled reconstruction. Do not output a second duplicate summary.

## Final image data

Use these exact headings. Put each field's value in its own code block.

1. `### 🔵 TITLE` — five to fifteen words.
2. `### 🟢 DESCRIPTION` — approximately 50 words explaining what happens, who is present, and why the moment matters.
3. `### 🟣 PASSAGE` — primary passage first, followed by relevant supporting references.
4. `### 🟠 KEYWORDS` — five to twenty simple, search-friendly, historically meaningful comma-separated terms; omit Bible references.
5. `### 🟤 HISTORICAL CONTEXT` — approximately 200 words explaining historical choices, reconstruction decisions, uncertainties, material culture, environmental realism, and why key choices were made.
6. `### 🟡 HISTORICAL CONFIDENCE NOTES` — three to five material elements labeled `High`, `Medium`, or `Low`, each with a short reason.
7. `### 🌅 IMAGE PROMPT` — normally 150–400 words, optimized for the requested provider. If none is named, default to Nano Banana Pro while retaining a model-neutral master direction in the project record.

The prompt should describe observational documentary realism, visible staging, environment, regional people, clothing and materials, natural posture, textures, practical objects, wear, construction, lighting, scale, and spatial relationships. Use natural light unless the scene requires another historically plausible source. People should look ordinary, regionally plausible, physically believable, and naturally posed—not Westernized, glamorous, airbrushed, or idealized. For ancient Jewish characters, use a plausible Levantine range of features and olive-to-brown complexions without making every person identical.

Include only relevant exclusions: anachronisms, modern materials, Renaissance conventions, fantasy, halos, theatrical glow, pristine sets, or fake writing. Do not overload the positive prompt with a long generic negative list.

Silently verify that all seven fields are present. End with one compact line offering: generate the image, provide JSON, create variations, or create a character profile. Do not generate an image unless the user explicitly requests it.

SHA-256: e02ca3bca84750b836084e3775c985685c2a20c6e9621b6f9fb4cf4a97b57c6f