--- name: blender-asset description: Produce one post-core Blender asset batch with source, render, export, manifest, and runtime evidence. disable-model-invocation: true --- Require: accepted core `artifact:` identity, frozen gameplay baseline, approved concept-art evidence, one bounded asset job, target workspace, and parent frame journal. Read the journal and open one job under the Blender-specific standard and applied `blender-bpy` method: tickets.py do --standard blender-game-asset --skill blender-bpy --parent --goal-file --isolation required The job fixes its source and input hashes, seed, units, axes, pivot, allowlists, budgets, cameras, render/export settings, and expected outputs. One fresh Blender process and directory preserve stdout, stderr, job, and partial output inventory. A successful handoff contains the `.blend` source, rendered previews, GLB export, asset manifest, hashes, zero-error Khronos validation, and a pinned production `GLTFLoader` scale/material/animation/collider probe. Invoke `review-delivery` in this existing asset-stage frame with the fixed batch, original job and evidence, browser-game-3d-asset and blender-game-asset pins, workspace, `workspace-adapter` git, caller `context-file`, `isolation` required, bound and outside asset probe. This independent batch owns its selected rounds and carries `repair-skill` blender-bpy; repair verification never resets them. Failure promotes nothing and records the exact diagnosis. A gameplay-altering scale, collider, animation, timing, camera, or readability change invalidates the affected core evidence and returns that seam to the gameplay gate. Missing Blender or image capability records `unverified` and parks only dependent asset work while preserving viable core evidence. Never install system software, purchase or accept unapproved assets, fabricate renders or manifests, overwrite a source authority, or call cosmetic output a runtime-qualified asset. Return: completed ticket carrying `artifact:` for the asset batch, source/render and export identities, manifest and runtime evidence, independent findings, invalidation set, and declared gaps.