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

blender-use

Skill4 starsChanged 21 days agoApache-2.0View on GitHub

Route requests to managed or Connector Blender workflows when a user wants Codex to create, modify, review, save, or export a Blender design.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 6 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Use this as the entry point.

    line 8
  2. Distinctive:

    Add-on, or a session exists merely because this Skill was loaded.

    line 16
  3. Distinctive:

    After the implementation brief, load only the required domain Skills:

    line 18
  4. Very specific:

    When the user asks for an end-to-end automated design, offer or honor one auto_with_budget envelope instead of requesting approval at every …

    line 42
  5. Distinctive:

    Before designing, turn the request into an implementation brief: scene purpose, required assets, object count and constraints, scale, …

    line 56
  6. Very specific:

    Before promising a production method, query capability.list for its domain and follow nextOffset until complete.

    line 60
  7. Very specific:

    If these catalog commands are absent in an older installed build, inspect the legacy session.capabilities and state the evidence gap rather …

    line 67
  8. Distinctive:

    If a required asset is missing, ask the user to provide it by default.

    line 70
  9. Distinctive:

    After the Blender artifact is ready, choose exactly one explicit route:

    line 85
  10. Distinctive:

    If intent is ambiguous, describe these three outcomes and ask the user to choose.

    line 96

≈ “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 Design Router

Use this as the entry point. Route new or Codex-launched sessions to blender-managed, already-open Blender windows to blender-connector, design work to blender-design-design, read-only questions to blender-design-inspect, visual evidence to blender-design-preview, approved outputs to blender-design-export, and failures to blender-recover.

On first use, call blender_connection_status. If no guarded Harness session is connected, route to blender-design-mcp-setup and call blender_getting_started; do not assume that Blender, the trusted Add-on, or a session exists merely because this Skill was loaded.

After the implementation brief, load only the required domain Skills:

  • scene organization and approved imports → blender-design-scene-assembly
  • mesh/product work → blender-design-hard-surface; curves → blender-design-curves
  • UV/material work → blender-design-uv-material; procedural nodes → blender-design-procedural-modeling
  • rigs → blender-design-character-rigging; animation → blender-design-character-animation
  • camera work → blender-design-cinematography; acceptance measurements → blender-design-quality-validation
  • sculpt, hair, or simulation → blender-design-sculpt-surface, blender-design-hair, or blender-design-simulation
  • rendering/compositing → blender-design-render-compositing
  • shot-list blocking or timing previews, or a white-model reference video for downstream video generation → blender-previs
  • Grease Pencil, tracking, or VSE → blender-design-grease-pencil, blender-design-tracking, or blender-design-sequence-editing
  • job.* → blender-design-background-jobs; add render-compositing for EXPORT, RENDER_STILL, or RENDER_ANIMATION_FRAMES; add sequence-editing for COMPOSE_VIDEO; add simulation for BAKE_POINT_CACHES

For multi-domain or conditional requests, read the maintained Skill routing reference.

Inspect before mutation and carry the returned scene revision. Use one transaction per milestone and fresh camera/front/side/top evidence. Interactive mode waits for milestone review; automatic mode evaluates the evidence and commits within the already approved task. Deletion, overwrite, expert Python and external actions retain action-bound authorization. Automatic fresh-file exports use the approved root, format scope and committed snapshot.

Input and delivery contract

When the user asks for an end-to-end automated design, offer or honor one auto_with_budget envelope instead of requesting approval at every milestone: approved output root, requested deliverables, missing-asset policy, optional downstream budget, and permission for Blender-designed proxies. Complete all safe milestones automatically, retain fresh evidence internally, and return a single final artifact inventory. interactive remains available for users who want stage review; review_only makes no mutation or export.

Pass the selected mode into the actual launcher or Connector; a conversational promise is not an active policy. Check session.status.executionPolicy before design. Show work in the foreground window using the Codex panel and registered view/playback commands. Follow foreground control and recovery for command arguments, takeover handling and limits. Respond to the user in their language, including Chinese.

Before designing, turn the request into an implementation brief: scene purpose, required assets, object count and constraints, scale, environment, camera plan, animation beats, duration, and target deliverables. Identify every missing reference asset explicitly.

Before promising a production method, query capability.list for its domain and follow nextOffset until complete; use capability.describe with {"id":"<registered command>"} to check arguments, prerequisites, availability, and evidence. L0 domains have no callable implementation. L1 is only a basic interface; L2 adds a workflow; L3 requires real-project, visual, and delivery validation; L4 adds recovery and compatibility. unknown availability means prerequisites still need inspection, not that they passed. Report tool, Skill, and real-project coverage separately. Expert Python does not fill a supported-domain claim. If these catalog commands are absent in an older installed build, inspect the legacy session.capabilities and state the evidence gap rather than inventing commands.

If a required asset is missing, ask the user to provide it by default. Only create a substitute asset when the user explicitly asks Codex to design it; mark that asset as a Blender-designed proxy rather than a supplied reference. Never silently invent a supplied reference's identity, motion, appearance, or exact dimensions.

At completion, return a named artifact inventory with absolute paths, formats, checksums, validation evidence, scene revision, and known deviations from the brief. If the user already chose the delivery route, honor it without asking again. Otherwise ask whether to finish locally or hand the validated artifacts to a separately installed downstream-renderer workflow. Export alone does not authorize upload, login, quotation, submission or payment.

Downstream AI rendering platforms are outside this plugin's responsibility.

Delivery routes

After the Blender artifact is ready, choose exactly one explicit route:

  • preview_only: use blender-design-preview or the receipt adapter and stop with a verified local file.
  • jimeng_web: use blender-to-dreamina against an already enabled official uploader in the foreground Connector session; stop at JimengLinkReady.
  • downstream_seedance: hand the validated artifact inventory to the separately installed codex-dreamina-3d workflow. That workflow owns approval, paid submission, query-only recovery, and final download.

If intent is ambiguous, describe these three outcomes and ask the user to choose. Never silently turn a preview export into a web handoff or a paid generation.

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.