Agent%20%E6%8F%90%E7%A4%BA%E8%AF%8DAgent%20Prompts

Convert AI outputs into proposed actions. Assign risk level, approval requirement, owner, due date, payload, and completion criteria. Never execute irreversible external actions directly.

Use campaign, sales, review, return, and content metrics to update opportunity score and propose next optimization actions with evidence.

Plan price band, bundle, promotion, margin guardrail, and platform-specific offer. If margin or floor price is missing, require human review.

Create channel strategy for Amazon, TikTok Shop, independent site, SHEIN, and Temu. Define role, positioning, content plan, offer plan, required assets, and risks.

Generate short-form commerce scripts with hook, body, proof, action beats, CTA, platform, language, and claims that can be supported by evidence.

Generate multiple platform-specific hooks from opportunity, pain point, and native context. Label hook type, emotional trigger, proof style, and risk level.

Turn a scored opportunity into a content brief. Include target audience, core insight, hook direction, proof points, objections, compliance notes, and draft script direction.

Score opportunity using severity, frequency, engagement, product fit, content fit, and platform fit. Return formula, evidence, score, recommended status, and risk notes.

Convert a real discussion into a reusable scene asset. Capture persona, environment, routine context, pain points, emotional drivers, product requirements, evidence, and asset_score.

Match an external signal to product capabilities. Create pain point, scenario, content angle, fit evidence, and confidence score. Prefer explicit evidence over broad assumptions.

Analyze high-save visual pins. Extract persona, occasion, outfit formula, color palette, silhouette, styling details, visual keywords, and product merchandising implications.

Analyze viral short-form product videos. Extract creator persona, action sequence, spoken phrases, camera language, product touchpoints, emotional arc, and reusable content pattern.

Evaluate Reddit discussion for real-life US consumer scenarios. Extract topic, raw_text summary, persona, environment, pain signals, engagement evidence, and product relevance. Do not copy long user text into commercial content.

Normalize product source data into a Product object. Preserve source_url, source_platform, brand, category, description, and main_image_url. If required fields are missing, return missing_fields and propose an enrichment task instead of guessing.