MA
marp-presen-review
>
Install
mkdir -p .claude/skills/marp-presen-review && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/15889" && unzip -o skill.zip -d .claude/skills/marp-presen-review && rm skill.zipInstalls to .claude/skills/marp-presen-review
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Marp形式プレゼンテーションの専門レビュー・評価・改善スキル。 10軸スコアリング（100点満点）で品質を定量評価し、具体的な改善コード付きで提案する。 Use when the user asks to: - Review, evaluate, or critique a Marp presentation / プレゼンをレビュー / 評価 / 添削 - Improve, polish, or upgrade existing slides / スライドを改善 / ブラッシュアップ - Check slide quality, score slides, or audit a deck / スライドの品質チェック / 採点 - "このプレゼンどう？" "スライドを見てほしい" "プレゼン資料のフィードバック" Reads .md files with Marp frontmatter, scores across 10 dimensions, and outputs a detailed review report with concrete before/after code improvements. Works as a companion to the marp-slides skill (generation → review → improvement cycle).601 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
About this skill
Marp Presen Review — Professional Evaluation & Improvement
Workflow
Phase 1 — Read & Parse
- Read the target
.mdfile - Verify Marp frontmatter (
marp: true) - Split into individual slides (by
---separator) - Count slides, identify slide classes, catalog layout components used
Phase 2 — Evaluate (10-Axis Scoring)
Score each axis 1–10. See references/scoring-rubric.md for detailed criteria.
| # | Axis | Weight | Key Question |
|---|---|---|---|
| 1 | Message Clarity | x2 | Each slide has exactly ONE clear message? |
| 2 | Story Structure | x2 | Logical narrative arc from opening to CTA? |
| 3 | Visual Hierarchy | x1.5 | Priority grasped in 3 seconds? |
| 4 | Information Density | x1.5 | Within 5-bullet / 12-word limit? |
| 5 | Layout & Whitespace | x1 | Grids, alignment, margins professional? |
| 6 | Typography | x1 | Fonts, sizes, emphasis consistent? |
| 7 | Color & Contrast | x1 | Palette limited, meaningful, accessible? |
| 8 | Visual Variety | x1 | No 3+ consecutive same-type slides? |
| 9 | Slide Titles | x1 | Titles are messages, not topic labels? |
| 10 | Technical Quality | x1 | Valid frontmatter, correct classes, no broken HTML? |
Total: 130 raw → normalized to 100 points.
Phase 3 — Diagnose
For each issue, produce:
[Slide N] [Axis] Severity: HIGH/MEDIUM/LOW
Problem: (one sentence)
Before: (current code)
After: (improved code)
Rationale: (design principle reference)
See references/common-antipatterns.md for frequent issues and fixes.
See references/design-checklist.md for visual design verification.
See references/story-evaluation.md for narrative structure patterns.
Phase 4 — Report
Output format:
# Marp Presen Review Report
## Score Summary
| Axis | Score | Grade |
|------|-------|-------|
| ... | 8/10 | A |
**Total: XX / 100** — Grade: S/A/B/C/D
## Strengths (Top 3)
## Critical Issues (Must Fix)
## Improvement Suggestions (Should Fix)
## Minor Polish (Nice to Have)
Grades: S (90+), A (80–89), B (70–79), C (60–69), D (<60)
Phase 5 — Improve
Ask user: "自動改善しますか？" If yes:
- Apply all Critical + Improvement fixes
- Restructure narrative arc if story score < 7 (use
references/story-evaluation.md) - Write improved file (same path with
-improvedsuffix) - Show before/after score comparison
Integration with marp-slides
Evaluation criteria align with marp-slides quality rules:
- Same CSS theme compatibility (excel-theme)
- Same layout component vocabulary (
.columns,.card,.highlight, etc.) - Same content rules (1 idea/slide, bold key phrase, vary visual types)
Quick Modes
| Mode | Phases | Trigger |
|---|---|---|
| Full review | 1→2→3→4→5 | "レビューして" "評価して" |
| Score only | 1→2→4 | "採点だけ" "スコアだけ" |
| Auto-improve | 1→2→3→5 | "自動改善して" "直して" |