Agent Skills: Visual Prompting: Six Blender Levers

Visual Prompting: Six Blender Levers

Created: 2026-10-03 | Last Updated: 2026-10-03

Recall this reference with six Blender levers or Blender mental model. The older phrase four Blender levers points here too: the corrected framework has six levers.

Use these levers to shape a Blender scene, diagram, image, or animation. They are a flexible prompting aid. Use what matters to the task; infer reasonable defaults and ask only about gaps that would change the result. Do not turn every request into six questions.

| Lever | Prompt focus | |---|---| | 1. Why it exists | Purpose, audience, or learning goal. What should the viewer understand? | | 2. What's off limits | Constraints: facts to preserve, privacy, scope, output limits, or effects to avoid. | | 3. What exists | Objects, scene contents, labels, data, and relationships. | | 4. How it looks | Materials, style, color, lighting, and visual hierarchy. | | 5. What it does | Motion, behavior, timing, and what each change reveals. For a still image, omit motion. | | 6. How we see it | Camera, viewpoint, framing, scale, and label legibility. |

Concise template

Create [visual and output format].
Why it exists: [purpose or learning goal].
What's off limits: [constraints and facts to preserve].
What exists: [objects, labels, relationships, or scene contents].
How it looks: [materials, style, color, lighting].
What it does: [motion or behavior, if needed].
How we see it: [camera, viewpoint, framing].

When revising a result, name the lever to change and preserve the others where they still work. Common failures: polishing before checking facts, decorative motion that hides the concept, or a camera that makes labels unreadable.

Teaching example: JOIN fan-out

Create a short Blender animation showing why summing an order total after a one-to-many JOIN can overcount.

• Why it exists: Reveal that changing row grain can repeat a measure without creating new value.
• What's off limits: Use fictional rows only. Preserve the arithmetic and label each grain. Do not imply that the repeated total is new revenue.
• What exists: One order row, order_id=101, order_total=60; two item rows, (101, A, 20) and (101, B, 40). JOIN on order_id to show two joined rows, each carrying order_total=60. Label the source as one row per order and the result as one row per item.
• How it looks: Flat cards, high-contrast labels, and consistent color for the order total. Use simple lighting so the numbers remain clear.
• What it does: Split the order card into the two joined rows. Highlight the repeated 60, then show SUM(order_total)=120 beside the original order total 60. Show item amounts 20 + 40 = 60 as the valid sum at item grain for this example.
• How we see it: Keep a fixed, front-facing camera. Frame the source, JOIN result, and comparison together, with labels legible at the target output size.

The motion reveals the repetition; the grain labels explain it. Check the numbers and labels before judging the render's polish.

Verification and scope

Tier: decision record. Keep the six prompt concerns separate so each visual can vary without a shared rendering implementation. Revisit if application tests show confusion between levers. Source: contributor-provided prompting model; the rows above are fictional.

This is a prompting aid, not a claim of universal effectiveness. That claim cannot be verified by a command. Review date: 2026-10-03.

Verify the example with this read-only, in-memory command:

python3 -c 'import sqlite3; c=sqlite3.connect(":memory:"); r=c.execute("WITH orders(id,total) AS (VALUES(101,60)), items(id,amount) AS (VALUES(101,20),(101,40)) SELECT COUNT(*),SUM(total),SUM(amount) FROM orders JOIN items USING(id)").fetchone(); assert r==(2,120,60),r; print(r)'

Expected: (2, 120, 60). Checked: 2026-10-03. Verify the actual output's arithmetic, grain, and legibility separately.