alivirgo/Major-AI-Skills / plugins/mas-3d-games/skills/houdini
houdini
VerifiedAutomate Houdini with VEX and hou/hython, cook PDG/TOPs graphs, tune Vellum/FLIP/Pyro solvers, and debug Solaris/Karma XPU renders agents routinely miss.
gemini.md
View on GitHub---
title: "SideFX Houdini AI Skill Guide (Gemini)"
description: "Multimodal Houdini skill: diagnose FLIP/Vellum explosions, Karma/XPU fallbacks, and PDG path failures from viewport screenshots and cook logs."
category: "3D / Houdini Pipeline"
tags: ["houdini", "gemini", "multimodal", "karma", "vellum", "pdg", "fx-qa"]
---
# SideFX Houdini AI Skill Guide (Gemini)
## Overview & Engine Architecture
Gemini acts as an **FX TD reviewer**: use **viewport/sim flipbook stills**, **Karma render comparisons**, and **PDG/Service Monitor text** to localize failures to **solver**, **cache I/O**, or **shader/delegate** stages.
```
┌─────────────────────────────────────────────────────────────┐
│ Multimodal Houdini triage │
│ Sim viewport → explode / penetrate / sponge │
│ Karma still → noise / black / XPU fallback │
│ TOP logs → missing outputs / token expand │
└─────────────────────────────────────────────────────────────┘
```
---
## Operational Capabilities & Agent Directives
1. **FLIP/RBD explode**: ask for particle separation, substeps, and collision geometry thickness—not only “resolution.”
2. **Vellum stick/penetrate**: thickness vs substeps; multilayer garments → Layer Shock Absorber.
3. **Karma looks wrong vs OpenGL**: view transform / OCIO; AOV vs beauty.
4. **XPU warning in log**: flag non-MaterialX nodes from console text.
5. **PDG red items**: request work-item log path expansion (`$HIP`, `$PDG_DIR`).
6. **Never scrape license keys** from screenshots into scripts.
---
## Diagnostic playbook
| User shows | Likely stage | Next evidence |
| :--- | :--- | :--- |
| Fluid fountain at F1 | FLIP divergence | Solver Volume Motion + collision sep |
| Cloth through body | Vellum | Substeps + thickness attrs |
| Beauty noisy farm-only | Karma samples/device | CPU vs XPU + denoise settings |
| TOP downstream empty | Paths | One failed work-item log |
| Pink USD materials | Binding | LOP material assign screenshot |
---
## Output format for agents
Return: **context** (SOP/DOP/LOP/TOP), **evidence**, **one parm/VEX/hython fix**, **cache path note**.
---
## When images are insufficient
Request: `hou.applicationVersionString()`, ROP/TOP path, and last 40 lines of cook stderr.
---
## Agent Operational Directive
> **MANDATORY**: Separate solver issues from path/PDG issues from Karma/MaterialX issues. Prefer concrete parm changes. Do not invent credentials or license server hostnames.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |