القائمة
Use only when the selected HOI4 portrait provider is the Comfy Cloud MCP route.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Use only when the selected HOI4 portrait provider is a configured loopback ComfyUI installation.
Use only when the selected HOI4 portrait provider is an existing RunPod ComfyUI workspace.
Use when creating, sourcing, processing, converting, organizing, wiring, or documenting visual assets for a Hearts of Iron IV mod.
Use for the provider-neutral HOI4 portrait source, prompt, archive, validation, and runtime handoff contract.
Use when coordinating custom Codex subagents for HOI4 mod implementation, asset production, text and audio research, audits, active small patches, planning handoffs, or documentation work.
Use when implementing, updating, auditing, or documenting Hearts of Iron IV events and event chains, including news and report events, on_action hooks, localisation, pictures, registration hooks, and validation.
| name | hoi4-comfyui-cloud |
| description | Use only when the selected HOI4 portrait provider is the Comfy Cloud MCP route. |
Read hoi4-portrait-production first. This is the only provider-specific
portrait route that may be used for this project. Do not use another provider
skill or invent a fallback route.
Use the official Comfy Cloud MCP endpoint:
https://cloud.comfy.org/mcp
Keep authorization in the provider/client session. Never write an API key, OAuth token, account data, or cookie to the mod, installation lock, logs, prompts, screenshots, or output.
The canonical bootstrap commands are:
curl -fsSL https://raw.githubusercontent.com/Comfy-Org/comfy-cloud-mcp/main/install.sh | bash
irm https://raw.githubusercontent.com/Comfy-Org/comfy-cloud-mcp/main/install.ps1 | iex
Use the provider subscription required for custom model import. Import the
exact current LoRA filename from the upstream lock:
hoi4_portraits_flux2_klein_9b_lora_000002500.safetensors.
The source and processing graphs also use the upstream
adaptive_portrait_crop custom node. Confirm that the current Builder
environment has that node before opening the graph. The processing graph does
not use the LoRA; a source run still emits three candidate master/game pairs
for review.
Use the matching pinned .api.json graph. Upload the approved source through
the provider upload flow; a project filesystem path is not a valid
LoadImage.image value. Dry-run the graph, submit it, retain the exact
prompt_id, wait for that job, and retrieve both 832x1120 and 156x210
outputs. A source run emits three candidate master/game pairs; review and
select one before handing the pair to the project. Queue emptiness is not
proof of success.
If authorization, subscription, model import, or MCP access is missing, keep the exact deferred status and the source-based fallback pending. Do not claim styled completion from a queued job or a provider preview.