// Generate character selfies with time awareness, context adaptation, and shooting mode selection
| name | clawmate-companion |
| description | Generate character selfies with time awareness, context adaptation, and shooting mode selection |
Generate contextualized character selfies based on user requests. Must strictly follow two-step calling: First call clawmate_prepare_selfie to get reference package, then generate prompt based on the package, finally call clawmate_generate_selfie.
Initiate the two-step image generation flow when user expresses these intents:
send a pic / send a selfie / show me a photo / 发张图 / 发张自拍what are you doing / where are you / 你在做什么 / 你在干嘛show me you at the coffee shop / take a pic in that outfit / 给我看你在咖啡店send another one / different expression / 再来一张 / 换个表情generate a character selfieclawmate_prepare_selfieExtract user intent keywords and call the tool to get reference package.
clawmate_prepare_selfie({
mode: "mirror" | "direct", // Required
scene?: string, // User-specified scene
action?: string, // User-specified action
emotion?: string, // User-specified emotion
details?: string, // Other details
})
Mode selection rules:
direct (DEFAULT): Use for all cases UNLESS user explicitly mentions mirror/full-body keywords belowmirror (SPECIAL CASE): Use ONLY when user explicitly says:
Critical: When in doubt, always choose direct. Do not infer mirror mode from context.
Return format (PrepareResult):
{
"timeContext": {
"period": "work",
"recommendedScene": "...",
"recommendedOutfit": "...",
"recommendedLighting": "..."
},
"modeGuide": {
"mode": "direct",
"requirements": ["phone not visible in frame", "..."]
},
"promptGuide": {
"requiredFields": ["scene", "action", "expression", "outfit", "lighting", "camera", "realism"],
"rules": ["single scene only", "..."],
"wordRange": "50-80 english words",
"example": "Photorealistic direct selfie, ..."
}
}
clawmate_generate_selfieYour role switches to image generation prompt engineer. This prompt's consumer is the image generation model, not humans.
Core principles:
clawmate_generate_selfie automatically attaches character reference images to the generation API. Reference images already carry character identity, so prohibit describing identity features (age, race, beauty) in the promptmodeGuide.requirements item must have corresponding description in prompt, cannot be omittedPrompt generation requirements:
promptGuide.requiredFieldstimeContext is only default recommendation; when user explicitly specifies time, scene, or outfit, prioritize user intentmodeGuide.requirements item (e.g., "direct eye contact to camera" must be written in)promptGuide.wordRangeclawmate_generate_selfie({
prompt: "<your generated complete English prompt>",
mode: "mirror" | "direct", // Keep consistent with Step 1
})
Calling example:
// Step 1
clawmate_prepare_selfie({ mode: "direct", emotion: "relaxed" })
// Step 2 (call after generating prompt based on returned package)
clawmate_generate_selfie({
prompt: "Photorealistic direct selfie, studying at a university library desk in the afternoon, open laptop and coffee cup in background, wearing comfortable hoodie, soft window light with warm ambient fill, focused but relaxed expression, medium close-up framing, natural skin texture, candid daily-life photo style",
mode: "direct"
})
On success (ok: true):
"Here you go~")imagePathOn failure (ok: false):
message in the return to continue conversationclawmate_generate_selfie without calling clawmate_prepare_selfie