anbeime/skill / skills/img2threejs/img2threejs
img2threejs
VerifiedTurn an object or character reference image into a quality-gated, animation-ready procedural Three.js model built in code. Use for image-to-3D reconstruction, detail-accurate object rebuilds, stylized/likeness-maximized human characters, sculpt specs, and staged code generation.
docs/standard-prompts/README.md
View on GitHub# Standard prompts
Four prompts, one per job. Each is copy-paste, each completes in one pass, and each states its own
exit condition. Replace every `<PLACEHOLDER>` before pasting.
| Prompt | Use when | Route |
|---|---|---|
| [`build.md`](build.md) | a reference image, no mesh | reconstruct from primitives, gated |
| [`glb-force-measured.md`](glb-force-measured.md) | a GLB is available and every parameter it measures should be forced, not approximated | measure, force, prove with parity gates |
| [`polish.md`](polish.md) | a build that renders but does not match | bounded correction, one defect at a time |
| [`vfx.md`](vfx.md) | the rig and its clips play, only the effects are missing | measure the events off a real mixer, then schedule against them |
Rigging and animation on top of a GLB: [`../GLB_ANIMATED_CHARACTER_PROMPT.md`](../GLB_ANIMATED_CHARACTER_PROMPT.md).
## Why these were rewritten
The previous prompts blocked more than they built, for three reasons that were mechanical rather
than stylistic:
1. **They pointed at a file that does not exist.** `GLB_ANIMATED_CHARACTER_PROMPT.md` delegated the
whole geometry route to `docs/GLB_CHARACTER_PROMPT.md`, which was never written. An agent
reaching the largest step of the run found a dead reference. `build.md` is that missing file.
2. **They ran a gate before producing the gate's input.** `validate_sculpt_spec.py --strict-quality`
fails with ~89 causes on a freshly scaffolded spec, because the scripts only scaffold —
`preSpecAssessment.objectClass.primaryType` is literally `"unassessed"`, all eight complexity
scores are `0`, and `detailInventory` is empty. Told to "run the gate and do not advance", an
agent stops on its first command, every time. **Every prompt here now fills the scaffold before
it validates**, and says exactly which fields.
3. **They treated every stop as equally serious.** A missing input and a failed measurement need
different responses. Each prompt below separates *produce the input and continue* from
*stop and ask*, and lists which is which.
## The rule the four share
> A gate is a question. Answer it with evidence, or say you could not — never by lowering the bar,
> and never by stopping before you have tried to produce the evidence it asked for.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| receiving-code-reviewobra/superpowersUse when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| requesting-code-reviewobra/superpowersUse when completing tasks, implementing major features, or before merging to verify work meets requirements | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| using-git-worktreesobra/superpowersUse when starting feature work that needs isolation from current workspace or before executing implementation plans - ensures an isolated workspace exists via native tools or git worktree fallback | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| verification-before-completionobra/superpowersUse when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |