agentleFS
full-aigc-pluginsfull-aigc-plugins/blender-design-plugin/skills/blender-previs/SKILL.md

blender-previs

Skill4 starsChanged 21 days agoApache-2.0View on GitHub

Use when turning a story or shot list into a color-coded white-model previs video in Blender, producing placeholder blocking, exact cut timing, and a machine-readable geometry-to-role map for downstream Seedance video generation.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 7 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Do not use this Skill for lookdev, final renders, or character performance.

    line 21
  2. Very specific:

    Read status first: Call blender_connection_status.

    line 26
  3. Distinctive:

    Add a ground plane and one neutral scale reference.

    line 40
  4. Distinctive:

    linear interpolation on every fcurve: default bezier AUTO handles let a neighboring key cliff (an explosion pop, a cut-boundary teleport) …

    line 42
  5. Very specific:

    Camera per shot: Follow blender-design-cinematography for placement, lens intent, and moves.

    line 47
  6. Distinctive:

    Never claim the previs represents limb motion, performance, or final visuals.

    line 82
  7. Distinctive:

    Never let two roles share one color, or change a role's color mid-project.

    line 83
  8. Distinctive:

    Never move a cut or retime a shot without updating the shot table first.

    line 84
  9. Distinctive:

    Never submit a paid video-generation request from this Skill.

    line 85
  10. Distinctive:

    Never treat a rendered previs as approval to skip the user's shot-table review.

    line 87

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Blender white-model previs

Previs answers three questions cheaply and deterministically: where the camera is, when the cut happens, and where every subject stands and moves. Iterate those here, in Blender, for free — never burn paid video-generation attempts on framing or timing guesses. This Skill produces a previs package, not a final-quality film: placeholders have no costume, texture, or limb articulation by design.

When to use

  • The user provides a story, script, or shot list and wants a blocking/timing preview.
  • A downstream video request (Seedance/Dreamina) needs a camera-and-timing reference video instead of text-only prompting.
  • An existing previs shot is rejected and must be re-blocked without touching the others.

Do not use this Skill for lookdev, final renders, or character performance. Route those to blender-design-render-compositing and blender-design-character-animation.

Workflow

  1. Read status first. Call blender_connection_status. With no guarded Harness session connected, route to blender-design-mcp-setup; never assume Blender exists because this Skill was loaded.
  2. Fix the shot table. Accept a structured shot table conforming to previs-shot-table.schema.json, or derive one from prose and get the derivation confirmed before building anything. Every shot needs a stable id, integer frame range at the project fps, shot size, camera intent, and per-subject position or motion beats. Frame arithmetic is the contract: the shot durations must sum to the project total exactly.
  3. Cast the placeholder scene. One primitive per role, cast from the default convention table below unless the user overrides it. Colors must stay visually distinguishable from each other and from the environment; one role never shares a color with another. Name every placeholder previs_role_<roleId> and record the full mapping as previs-map.schema.json content. Add a ground plane and one neutral scale reference.
  4. Block, do not perform. Keyframe only object location and orientation, with linear interpolation on every fcurve: default bezier AUTO handles let a neighboring key cliff (an explosion pop, a cut-boundary teleport) bend unrelated segments and drag a tracked subject off-center. Limb articulation, facial performance, and physics are intentionally absent; say so in the deliverable instead of approximating them.
  5. Camera per shot. Follow blender-design-cinematography for placement, lens intent, and moves. A search/POV shot aims at what the character is looking for — key the tracking target onto the subject being found, not at empty space. Cuts are hard frame-range boundaries from the shot table; add a transition only when the table explicitly requests one.
  6. Render cheap, per shot. Render each shot as its own frame-range job (job.submit with RENDER_ANIMATION_FRAMES, Workbench or low-sample EEVEE) so a rejected shot re-renders alone. Via blender-design-background-jobs.
  7. Stitch with exact cut frames. Assemble the per-shot clips in the sequencer (COMPOSE_VIDEO, via blender-design-sequence-editing) and verify the composited duration equals the shot-table total in integer frames.
  8. Deliver the previs package: the previs video, the shot table, the geometry-to-role map, and the downstream handoff prompt drafted from seedance-handoff-prompt.md — that prompt is what codex-dreamina-3d-from-blender or dreamina-video-production consumes to turn this previs into a real shot.
  9. Iterate in place. A rejected shot keeps its id; re-block and re-render only that shot, restitch, and reissue the package. Never renumber shots: downstream references and approvals hang off the ids.

Default color convention

Override as a set, per project, in the geometry-to-role map; never ad hoc per shot.

Role classPrimitiveColor
ProtagonistCubeLight cyan #7FD4C1
Antagonists / pursuersCubes, one eachDark gray #4A4A4A
Hero prop (rocket, weapon, macguffin)CylinderRed #D9483B
Ground vehicleBoxWhite #F2F2F2
Rail or large vehicleLong boxTeal #2E8B8B
Environment landmarkKeep real blocking mass, untexturedNeutral gray

Never do

  • Never claim the previs represents limb motion, performance, or final visuals.
  • Never let two roles share one color, or change a role's color mid-project.
  • Never move a cut or retime a shot without updating the shot table first.
  • Never submit a paid video-generation request from this Skill; producing the handoff prompt is where it ends.
  • Never treat a rendered previs as approval to skip the user's shot-table review.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.