
slide
by lexoliu
Providing unified trait abstractions for AI models
SKILL.md
name: slide description: "Create Markdown-based presentation slides using Slidev. Use when users ask to create slides, presentations, or slide decks."
Slide Skill (Slidev)
Pre-flight Check
Ask: "Will you edit this in Keynote/PowerPoint later?"
- YES: Stop. Recommend using those apps directly.
- NO: Continue.
Workflow
Phase 1: Research (Optional)
If the topic needs research, use the builtin research subagent:
task research --prompt "Research [TOPIC] for a presentation. Find key concepts, recent statistics, and expert insights."
Use explicit flags to avoid positional argument ambiguity:
task --subagent research --prompt "Research [TOPIC] for a presentation. Find key concepts, recent statistics, and expert insights."
Phase 2: Requirements (Main Agent)
Gather from user:
- Topic and key message
- Audience: technical / general / executive
- Slide count (default: 5-8)
- Style: dark / light / minimal / corporate / playful
Phase 3: Art Direction (Subagent)
Launch art direction subagent to create design guide and outline:
task --subagent .skills/slide/subagents/art_direction.md --prompt "Create design guide for: [TOPIC]. Audience: [AUDIENCE]. Style: [STYLE]. Slide count: [N]."
The subagent returns a YAML design guide containing:
- Theme selection
- Color palette
- Typography guidance
- Slide-by-slide outline with layouts and content
Phase 4: User Approval
Present the outline to the user:
- Show slide titles and structure
- Confirm design direction
- Get approval before proceeding
Phase 5: Parallel Slide Creation (Subagents)
Launch one slide_creator subagent per slide in parallel:
task --subagent .skills/slide/subagents/slide_creator.md --prompt "Slide 1: [SPEC]. Design: [DESIGN_GUIDE]"
task --subagent .skills/slide/subagents/slide_creator.md --prompt "Slide 2: [SPEC]. Design: [DESIGN_GUIDE]"
task --subagent .skills/slide/subagents/slide_creator.md --prompt "Slide 3: [SPEC]. Design: [DESIGN_GUIDE]"
# ... one per slide
Each subagent:
- Creates standalone temp file with slide content
- Renders to PNG for visual verification
- Views the image to ensure quality
- Iterates until the slide looks correct
- Returns verified slide markdown
Phase 6: Assembly & Preview (Main Agent)
Collect all slide outputs and assemble into final presentation:
mkdir -p artifacts
cat > artifacts/slides.md << 'EOF'
---
theme: [from design guide]
title: [presentation title]
transition: [from design guide]
---
[Slide 1 content]
---
[Slide 2 content]
---
[... all slides ...]
EOF
Clean up temp files and start preview server:
rm -rf artifacts/temp
cd artifacts && npx slidev slides.md --port 3030 &
Tell user: "Preview at http://localhost:3030"
Phase 7: Export (Optional)
If user wants PDF:
cd artifacts && npx slidev export slides.md --output slides.pdf
Reference Files
Read these from the skill directory:
.skills/slide/references/syntax.md- Full Slidev syntax guide.skills/slide/references/layouts.md- Layout options and usage.skills/slide/references/themes.md- Theme gallery and recommendations
スコア
総合スコア
リポジトリの品質指標に基づく評価
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
レビュー
レビュー機能は近日公開予定です