vickyhan924 avatar

Douyin Viral Content

This is a self-optimizing Douyin copywriting generator that won't give you output until it hits 5-star quality. You feed it a quote or excerpt, and it automatically pulls historical performance data from your published posts, applies nine viral content factors (emotional reversal, cognitive reframing, authority markers), scores itself across 300 points, and iterates until it clears the threshold. It learns from what actually worked for your account, adjusts prediction models if it's been over or underestimating views, and outputs polished copy with hashtags and view predictions. The whole thing runs as one command with no manual intervention. Good if you're running a Chinese short video account and want consistent quality without babysitting each draft, though the rigid quality gate means you're trusting its scoring system completely.

skills.sh installs
1.7K
GitHub stars
3
last updated
Aug 21, 2026

Install

$ npx -y skills add vickyhan924/self-media-script --skill douyin-viral-content --agent claude-code

Installs into .claude/skills of the current project.

Files

抖音爆款文案生成器

核心理念

用户只需说:"生成新文案",提供书籍摘录或金句内容。

所有分析、优化、打分、预估全部自动完成,直到输出5星级爆款文案。

全自动化工作流程

当接收到生成文案请求时,按照以下步骤自动执行:

步骤1:历史数据自动读取与分析

自动扫描 已发布/ 文件夹下的所有历史文案:

  • 提取每条文案的预估播放量、实际播放量、各维度得分
  • 识别哪些爆款要素得分高的文案实际表现更好
  • 分析预估模型是否存在系统性偏差(总是高估或低估)
  • 总结账号的内容特色(更擅长哪类情感、风格、话题)

如果历史数据不足(少于3条):使用通用模型,无需调整策略

步骤2:策略自动校准

基于历史分析结果,自动调整:

  • 要素权重:提高表现好的要素权重(如"情感权力反转"高分文案播放量好,则增强该要素)
  • 预估参数:校准播放量预估模型(如历史总是高估20%,自动降低预估系数)
  • 标签策略:优化标签选择(如某些标签组合表现更好,优先使用)
  • 内容方向:强化账号擅长的内容类型

步骤3:应用9大爆款要素生成文案

参考 references/viral-factors.md,确保文案包含关键要素:

  1. 情感权力反转
  2. 认知重构
  3. 权威背书
  4. 宿命论哲学
  5. 算法友好标签
  6. 金句传播性
  7. 完播率优化
  8. 互动钩子设计
  9. BGM适配建议

融入历史学习:基于步骤2的策略调整,强化成功要素。

步骤4:多维度打分评估

参考 references/scoring-system.md,自动计算:

  • 内容质量分(100分):爆款要素覆盖度、金句质量、情感共鸣度
  • 算法适配分(100分):标签策略、完播率预期、互动引导
  • 创新度分(100分):角度创新性、差异化表达
  • 综合爆款指数(总分300,换算成5星制)

步骤5:质量控制与自动优化迭代

质量标准:必须达到5星(240-300分)才能输出

如果首次生成未达到5星:

  1. 识别得分最低的维度(如创新度不足)
  2. 自动优化该维度(如:更换角度、增强金句、优化开头)
  3. 重新打分评估
  4. 重复迭代直到达到5星标准

自动规避风险:在生成过程中自动规避潜在风险(如文案过长、开头不够吸引、标签不够正向等),无需告知用户。

自动应用优化点:在迭代过程中直接优化文案,无需告知用户优化了什么。

步骤6:播放量智能预估

Sponsors

Details

Repository
github.com/vickyhan924/self-media-script
Category
Automation & Workflows
1huggingface avatarhuggingface-tool-builderhuggingface/skillsUse this skill when the user wants to build tool/scripts or achieve a task where using data from the Hugging Face API would help. This is especially useful when1.7Kskills.sh installs
2nvidia avatartao-finetune-huggingface-modelnvidia/skillsFine-tune any HuggingFace CV / VLM / LLM model on local NVIDIA GPUs inside an NGC PyTorch container. Use when the user wants to fine-tune a HuggingFace model (f1.7Kskills.sh installs
3nodeops-app avatarcreateosnodeops-app/skillsDeploy ANYTHING to production on CreateOS cloud platform. Use this skill when deploying, hosting, or shipping: (1) AI agents and multi-agent systems, (2) Backen1.7Kskills.sh installs
4loops-so avatarloops-apiloops-so/skillsUse this skill whenever the user wants to integrate Loops from application code, backend services, webhook handlers, or server-side automation. This includes th1.6Kskills.sh installs
5lingzhi227 avatarpaper-assemblylingzhi227/agent-research-skillsOrchestrate the full paper pipeline end-to-end. Manage state propagation between phases (literature → plan → code → experiments → figures → tables → writing → r1.6Kskills.sh installs
6K-Dense-AI avatarbulk-rnaseqK-Dense-AI/scientific-agent-skillsEnd-to-end bulk RNA-seq orchestrator — takes raw FASTQ reads through QC and trimming (FastQC, fastp/Trim Galore), alignment and quantification (STAR, Salmon, fe1.5Kskills.sh installs