"Guides Krita AI Diffusion WorkflowInput construction, prompt
Scanned 9/8/2026
Install to Claude Code
npx -y skills add VectorSpaceLab/AREX-Skill --skill inference-workflows --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Inference Workflows?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/vectorspacelab-inference-workflows)More formats (shields.io, HTML) on the badges page.
---
name: inference-workflows
description: "Guides Krita AI Diffusion WorkflowInput construction, prompt
preparation, inpaint, refine, upscale, control-image, and ComfyUI workflow
lowering tasks."
disable-model-invocation: true
metadata:
disco-role: operating
license: GPL 3.0
---
# inference-workflows
Use this sub-skill when the task is about image-generation request payloads or
how Krita AI Diffusion lowers plugin state into local ComfyUI/cloud work.
Trigger examples:
- Building, inspecting, serializing, or deserializing `WorkflowInput`.
- Choosing among `WorkflowKind.generate`, `refine`, `inpaint`, `refine_region`,
`upscale_simple`, `upscale_tiled`, `control_image`, and `custom`.
- Debugging prompt/style/LoRA/wildcard/layer-token preparation before a job is
sent to a client.
- Explaining inpaint fill/mask/reference choices or selection/crop extents.
- Estimating `passes_count`, `cost`, or request image payload shape without
running actual generation.
- Reviewing ComfyUI workflow lowering in `backend.workflow` or
`backend.comfy_workflow`.
## Safe entry points
```bash
python sub-skills/inference-workflows/scripts/inspect_workflow_input.py --kind generate --round-trip
python sub-skills/inference-workflows/scripts/inspect_workflow_input.py --kind inpaint --round-trip
python sub-skills/inference-workflows/scripts/inspect_workflow_input.py --kind upscale-tiled --target 2048x1536 --round-trip
```
These commands construct tiny in-memory payloads and do not connect to ComfyUI,
launch Krita, download models, call cloud APIs, or generate images.
## References
- [references/api-reference.md](references/api-reference.md): dataclass/API
contracts for `WorkflowInput` and related inputs.
- [references/workflow-recipes.md](references/workflow-recipes.md): request
recipes for generate, refine, inpaint, upscale, control image, prompt/LoRA,
and ComfyUI lowering.
- [references/troubleshooting.md](references/troubleshooting.md): workflow-kind,
prompt, mask, model, and serialization failure recovery.
- [scripts/inspect_workflow_input.py](scripts/inspect_workflow_input.py): bundled
offline request inspector.
## Boundaries
- For server URLs, model catalogs, custom nodes, cloud/local clients, or managed
server errors, route to `server-resources`.
- For Qt workspace state that decides when `DocumentModel.generate()` enqueues a
particular workflow kind, route to `ui-workspaces` as well.
- For custom ComfyUI Graph workspace metadata and ETN placeholders, route to
`custom-graphs`.
- For `Image`, `Mask`, `Bounds`, layer tokens, prompt comments, style JSON, or
PNG metadata, route to `document-image-state`.

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!