菜单
Generate podcast cover art with AI and apply branding. Uses Gemini via OpenRouter for image generation with light cream backgrounds, then adds Yudame Research logo and series/episode text using Playfair Display typography. Requires OPENROUTER_API_KEY.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | podcast-cover-art |
| description | Generate podcast cover art with AI and apply branding. Uses Gemini via OpenRouter for image generation with light cream backgrounds, then adds Yudame Research logo and series/episode text using Playfair Display typography. Requires OPENROUTER_API_KEY. |
Skill name: podcast-cover-art
Generate episode cover art using AI and apply podcast branding in one command.
cd ~/src/research/podcast/tools
python cover_art.py ../episodes/YYYY-MM-DD-slug/
This:
content_plan.mdreport.md using Gemini via OpenRoutercover.png in episode directory# Full generation + branding (default)
python cover_art.py ../episodes/your-episode/
# With explicit series name
python cover_art.py ../episodes/your-episode/ --series "Series Name"
# Custom episode text for overlay
python cover_art.py ../episodes/your-episode/ --episode-text "Ep 5 - Topic"
# Skip AI generation, only apply branding to existing cover.png
python cover_art.py ../episodes/your-episode/ --skip-generate
# Quiet mode
python cover_art.py ../episodes/your-episode/ --quiet
report.md must exist (used to generate AI image prompt)content_plan.md should exist (for auto-detecting title/series)OPENROUTER_API_KEY in environment (for Gemini image generation)# Verify fonts installed
cd ~/src/research/podcast/tools
uv run python add_logo_watermark.py --check-fonts
# If missing, install:
mkdir -p ~/Library/Fonts && cd ~/Library/Fonts
curl -L -o playfair.zip "https://gwfh.mranftl.com/api/fonts/playfair-display?download=zip&subsets=latin&variants=600,italic"
unzip -o playfair.zip
cover.png - Final branded cover art (1024x1024, ~1MB)logs/cover_generation_*.txt - Generation loglogs/cover_generation_*.json - Prompt metadatalogs/branding_log_*.txt - Branding log| Property | Value |
|---|---|
| Dimensions | 1024x1024px |
| Background | Light cream (#F5F1E8) |
| Accents | Black (#000000), Salmon (#E8B4A8) |
| Typography | Playfair Display |
| Branding | Yudame logo + series/episode text |
| Format | PNG, ~500KB-1MB |
Cover art runs during the publishing phase (Phase 12) alongside feed.xml updates:
Phase 12: Publishing
├── Generate cover art (can run in parallel with metadata)
├── Create logs/metadata.md
├── Update feed.xml
└── Validate with podcast-feed-validator
This allows cover art to generate while other publishing tasks proceed.
Validate podcast RSS feed against specification standards. Checks channel metadata, episode elements, XML structure, and file metadata accuracy. Use after updating feed.xml with a new episode to ensure compliance with RSS spec. References docs/RSS-specification.md for requirements.
Evaluate completed podcast episodes across 10 quality dimensions. Diagnostic tool that produces detailed scorecards with evidence-based ratings, strengths, weaknesses, and workflow improvement recommendations.
Automate Perplexity Deep Research API calls using sonar-deep-research model. Supports sync (blocking) and async (fire-and-poll) modes. Use for Phase 1 academic research in podcast episodes. Returns research ready to paste into research-results.md.
DEPRECATED - Use gpt-researcher skill instead. This browser automation approach has been replaced with the local GPT-Researcher framework using OpenAI o1.
LEGACY SKILL - Manual NotebookLM web interface workflow. The primary workflow now uses notebooklm-enterprise-api for automated audio generation. Use this skill only when the API is unavailable.
Transform research materials into structured episode plans that guide NotebookLM audio generation. Creates content_plan.md with three-section structure, Wave 2 structural design, and NotebookLM guidance.