Feishu PPT Skill — Lark Slides Template Library for AI Agents

AI-agent skill for building Lark (Feishu) slides via the lark-cli: a 51-page template library, brand design tokens, XML generation workflow, and automated layout review.

Built as a reusable SKILL.md so agents can produce polished, on-brand presentations quickly — pick a template, swap content, lint, publish.

What's inside

.
├── SKILL.md                 # Design spec + template workflow + CLI commands + troubleshooting
├── templates/
│   ├── INDEX.md             # 51-page scene index (per page: use case / structure / replace list)
│   └── slide01.xml ~ slide51.xml  # 960×540 Lark Slides XML templates
├── assets/
│   ├── cherry-logo.png      # Example logo asset (1024×1024 transparent PNG)
│   └── product-placeholder.png  # Product screenshot placeholder
└── scripts/
    └── review_layout.py     # Automated layout check (overlap / overflow / out-of-bounds)

Template coverage

Quick start (agent view)

1. Read templates/INDEX.md, pick a template by content type
2. Copy slideXX.xml + assets/*.png into a work dir
3. Replace sample content per the INDEX "replace" checklist (templates are a starting point — adjust layout/density to fit content, keep the brand tokens)
4. Lint: python3 scripts/review_layout.py --input slide.xml
5. Publish via lark-cli slides +create / +add-slide

Design tokens

• White canvas #FFFFFF + bold black headings #171717 + white cards with thin border #D6D6D2
• Coral red #FF5A5F as the single brand accent (logo #FF5757)
• Black button = the only CTA; restrained accent lines (pink/cyan/purple/blue/green/yellow)
• 17 anti-patterns (no blue-purple gradients / no coral-red fills / no heavy shadows-glassmorphism...)
• Exact palette & font sizes: see SKILL.md (template values are authoritative)

Dependencies

• larksuite/cli (Lark Slides XML operations)
• Official xml_lint.py (inside the lark-slides skill) for XML validation

License

• Code, templates, and scripts: MIT © 2026
• Brand assets (e.g. assets/cherry-logo.png): brand/logo assets are not covered by the MIT license — they are the property of their respective owners and included only as examples. Replace them with your own assets for production use.