Jimeng (即梦) Runtime Skill

Use this skill to generate/edit media via Jimeng (即梦) through the Jimeng MCP (mcp__jimeng__*). Jimeng handles Chinese and English prompts and is a strong default for Chinese-language creative requests.

When To Use

Use Jimeng when the user asks to generate an image, edit/transform an existing image, or generate a short video — particularly with Chinese prompts or for a China-market aesthetic. For English-first or specific Replicate models use mcp_replicate.

Connection

Authenticates with Jimeng credentials. On an auth error, stop and ask the user to reconnect. Generation is billable / not instant — treat each call as a paid run.

Core Tools

• generate_image — prompt (中文 or English) → one or more images.
• edit_image — transform an existing image with a text instruction.
• generate_video — prompt → short video.

Common Recipes

Generate an image

1. Confirm the prompt (and count/aspect if relevant). 2. generate_image. 3. Return the asset reference.

Edit an image

1. Take the source image + the edit instruction. 2. edit_image. 3. Return the result.

Guardrails

• Each generation is billable and not instant — confirm the prompt before generating; don't fan out many variations unless asked.
• For edit_image, confirm which source image is being transformed so you don't edit the wrong asset.
• Respect content policy and platform rules (no disallowed or infringing content).
• Don't loop-retry a slow/failed generation without asking — re-runs cost again.

Edge Cases & Errors

• Output is an async asset reference/URL — return it; don't fabricate a result before completion.
• Video is slower/pricier than image — set expectations and confirm first.
• Auth/quota errors → stop and tell the user.