agentleFS
holon-runholon-run/holon/skills/3d-production-workflow/SKILL.md

3d-production-workflow

Skill194 starsChanged 4 days agoApache-2.0View on GitHub

Turn a creative brief into shared assets, editable scenes and verified 3D games or animated films. Use for cross-stage production, asset reuse and multi-view correction, not engine API reference or ordinary video editing.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep a small production manifest: stable asset ID, purpose, reuse, provenance/license status, creation method, scale/axes, material style, …

    line 18
  2. Distinctive:

    Prefer procedural geometry for simple platforms, crystals and collision shapes.

    line 22
  3. Distinctive:

    If references are absent, make a composition sketch or graybox and record design choices.

    line 25
  4. Distinctive:

    Test the delivery path with a minimal sample before expensive production: authentication, accepted formats/MIME, CSP, size limits, …

    line 28
  5. Very specific:

    Check front/side/back or overhead views plus the actual output camera for silhouette, occlusion, proportion, intended floating/ground …

    line 62
  6. Distinctive:

    Film: use its own camera/timeline, deterministic fixed-timestep frame rendering and video encoding.

    line 77
  7. Very specific:

    Keep build/run/render commands, lockfiles, asset manifest, actual cost and evidence.

    line 83

≈ “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

3D Production Workflow

Move from a generated model to a playable work or deliverable film. Use the project's chosen engine and technical skills; this workflow does not install tools, grant spending permission or establish production readiness.

Decompose the work before generating assets

  • List subjects, environment, interactions, cameras and motion from the target frame or gameplay. With reference images, distinguish occlusion, perspective, silhouette and lighting; a hidden back surface is unknown, not established geometry.
  • Keep a small production manifest: stable asset ID, purpose, reuse, provenance/license status, creation method, scale/axes, material style, performance limits and acceptance viewpoints.
  • Prove space, camera framing, reachability and gameplay with a graybox first. Prefer procedural geometry for simple platforms, crystals and collision shapes. Consider generated/external assets when complex silhouettes matter to recognition, not for every object.
  • If references are absent, make a composition sketch or graybox and record design choices. Do not buy or upload material merely to satisfy a reference step.
  • Test the delivery path with a minimal sample before expensive production: authentication, accepted formats/MIME, CSP, size limits, playback/download. Local rendering success is not hosted delivery. If unsupported, preserve the original files and report local-file delivery limits; do not bypass login, modify the host without approval, or call a frame preview MP4 playback.

Generation is not assembly

  • Confirm model/quality, expected cost, task count and retry limits before generation. Without spending approval, continue only with grayboxing and free steps.
  • For optional Tripo, use the configured UXC interface and local connection instructions. Discover operations, inspect schemas and verify current official pricing; do not guess parameters from tutorials or mix API versions. Require both UXC ok == true and Tripo data.code == 0.
  • Distinguish request acceptance, remote terminal success, download and scene import. Record the task ID before waiting. After uncertain submission, reconcile known tasks/usage before retrying a chargeable request.
  • Preserve original models and reproducible import settings. Record provenance, generation parameters, actual credits and license status, not keys or signed download URLs. Downloading from a third-party CDN must not forward the service API key.
  • Inspect readable meshes/textures, bounds, units, up/forward axes, origin, color space and triangle count. Normalize scale/orientation in an explicit scene adapter rather than repeatedly destructively editing the source.

Share scenes and correct multiple views

  • Reuse source assets through the manifest and rebuildable scene modules, not divergent game/film copies. Separate collision and navigation/reachable areas from visual geometry.
  • Establish subject scale/style before assembling environment, materials, lighting and cameras. Revisit the original intent each iteration rather than optimizing a single attractive viewpoint.
  • Check front/side/back or overhead views plus the actual output camera for silhouette, occlusion, proportion, intended floating/ground contact and intersections. Use real rendered screenshots; do not claim visual-model review unless it actually ran.
  • For characters only when needed: multi-view consistency, mesh, rig/skin, motion adaptation, then contact/intersection checks. Skip this branch for non-character or unrigged work. A generated model does not prove usable character animation.

Validate each requested output

  • Game: deliver source and a playable build with controls, objective and applicable win/loss/restart behavior. Run real builds and browser/WebGL checks for input, collision/collection, end/restart states, console and asset requests. A film cannot substitute for playable acceptance.
  • Film: use its own camera/timeline, deterministic fixed-timestep frame rendering and video encoding. Deliver editable scenes, render/encode entrypoints and the film. Check dimensions, frame rate, frame count/duration with ffprobe and inspect first/middle/last and key-action frames. An orbit preview or screen recording does not substitute for editable source and the agreed rendered film.
  • Keep build/run/render commands, lockfiles, asset manifest, actual cost and evidence. Separate assertions, screenshot checks and visual review; name the tested device/browser and untested delivery paths. If import fails, a graybox may validate other stages, but report the generated-asset path as incomplete.
  • Rebuild every affected output after shared-asset edits. Improve this workflow from observed failures rather than adding speculative instructions or generalizing one successful trial into a production-quality claim.

Method provenance

Inspired by Saccc_c's 3D production article: reference decomposition, differentiated asset creation, scene assembly, multi-view correction and an optional character branch. This is an original workflow synthesis, not a reproduction or endorsement. Game/film acceptance and cost controls are engineering additions, not claims about the article. A reference does not grant asset rights or execution authority.

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.