SKILL.md

$2a

Choose Charts: Use the Chart Selection Dictionary in references/VIS-GUIDE.md to pick the most appropriate visualization for each point. If the user provides data (tables/CSV), optionally call scripts/chartkit.py to generate PNG charts; otherwise, create placeholder diagrams with a list of required data fields.

Layout & Style: Follow references/STYLE-GUIDE.md for font sizes, line spacing, white space, contrast ratios, color palettes, and accessibility (WCAG AA compliance).

Speaker Notes: Generate 45-60 second speaker notes for each slide, structured as: opening → core assertion → evidence explanation → transition.

Self-Check & Score: Use references/CHECKLIST.md for a pre-flight check, then score with references/RUBRIC.md If total score < 75, identify the weakest 3 items and refine; repeat scoring (max 2 iterations).

Deliverables (all saved to /output/):

/output/slides.md: Markdown slides (Marp/Reveal.js compatible), with assertion-style headings + bullet points/chart placeholders + notes

/output/assets/*.png: Generated charts (if applicable)

/output/notes.md: Full speaker notes and delivery outline

/output/refs.md: Citations and data sources

/output/presentation.pptx: If python-pptx is available, export to PPTX; otherwise, keep Markdown and include instructions for "one-click conversion to PPTX" in the first screen (does not block delivery)

Orchestration Mode (End-to-End Automation)

When the user requests a "complete" or "presentation-ready" deliverable, ppt-creator automatically orchestrates the full pipeline: content creation → data synthesis → chart generation → dual-path PPTX creation (Marp + document-skills:pptx) → chart insertion. This typically delivers TWO complete PPTX files with different styling for user comparison.

Activation: Phrases like "complete PPTX", "final deliverable", "ready for presentation"

Duration: 4-6 minutes (parallel execution)

Output: presentation_marp_with_charts.pptx + presentation_pptx_with_charts.pptx

For orchestration details, see references/ORCHESTRATION_OVERVIEW.md (start here), then navigate to specialized guides as needed.

Core Principles (Must Follow)

Information Organization: Conclusion first, then evidence (Pyramid Principle). Each slide conveys only 1 core idea. Headings must be testable assertion sentences, not topic labels.

Evidence-First: Use charts/tables/evidence blocks instead of long paragraphs; limit to 3-5 bullet points per slide.

Data Visualization: Chart selection and labeling (axes/units/sources) must comply with references/VIS-GUIDE.md If data is insufficient, provide "placeholder chart + list of missing fields."

Accessibility: Color and text contrast must meet AA standards (see STYLE-GUIDE). Provide alt/readable descriptions for charts and images.

Reusability: Use consistent naming, stable paths, reproducible output. Do not hard-code random numbers in code.

Safety & Dependencies: Do not scrape the web without permission. Only run scripts when user provides data. If matplotlib/pandas are unavailable, fall back to text + placeholder diagram instructions.

Workflow Overview

Stage 0 - Archive Input: Record user's original request, defaults used, and assumptions made.

Stage 1 - Structure Goals: Rewrite the goal into "who takes what action when" (clear CTA).

Stage 2 - Storyline: Use Pyramid Principle to define "one-sentence conclusion → 3-5 first-level reasons → evidence."

Stage 3 - Outline & Slide Titles: Create a 12-15 slide chapter skeleton. Each slide has one assertion-style heading.

Stage 4 - Evidence & Charts: Use the Chart Selection Dictionary from VIS-GUIDE. If data is provided, call chartkit.py to generate PNGs; otherwise, create placeholder + required field list.

Stage 5 - Layout & Accessibility: Apply STYLE-GUIDE for font sizes, spacing, contrast ratios, color palettes; unify units and decimal places.

Stage 6 - Speaker Notes: Generate 45-60 second notes per slide: opening → assertion → evidence explanation → transition.

Stage 7 - Self-Check & Scoring: Run CHECKLIST; score with RUBRIC. If score < 75, focus on weakest 3 items, refine, re-score (max 2 iterations).

Stage 8 - Package Deliverables: Generate /output/ directory with slides.md / notes.md / refs.md / assets/*.png. If python-pptx is available, export PPTX.

Stage 9 - Reuse Instructions: Append a "5-step guide to replace data/colors with your own" at the end of notes.md.

Resources

references/INTAKE.md

Minimal 10-Item Questionnaire (use defaults if missing):

Who is the audience? (Default: general public)

Core objective? (Default: "understand and accept" a proposition)

Desired action/decision? (Default: agree to move to next step after the meeting)

Duration & slide count limit? (Default: 15-20 min, 12-15 slides)

Tone & style? (Default: professional, clear, friendly)

Topic scope & boundaries? (Default: given topic + 1 layer related)

Must-include points/taboos? (Default: none)

Available data/tables? (Default: none; can generate structure placeholder + list required fields)

Brand & visual constraints? (Default: built-in neutral theme)

Deliverable format preference? (Default: slides.md + optional PNG charts; export PPTX if available)

references/WORKFLOW.md

Detailed step-by-step process from "topic" to "presentation-ready output."

references/TEMPLATES.md

Slide Template Library (assertion-evidence style):

Cover, Table of Contents, Problem Statement, Opportunity/Goal, Solution Overview, Evidence 1-3, Risk & Mitigation, Case Study/Comparison, Roadmap/Timeline, Conclusion & Actions, Backup Slides

Micro-templates: Comparison (A vs B), Pyramid Summary, Process 4-Step, KPI Dashboard, Geographic Distribution, Funnel, Pareto, Sensitivity, Cost Structure (Waterfall), Contribution (Stacked)

references/VIS-GUIDE.md

Data Visualization Selection & Labeling Standards:

Chart Selection Dictionary (common questions → chart types)

Labeling & units (axes, units, data scope, time range; source in footer)

Accessibility & contrast (WCAG 2.1 AA: text vs background ≥ 4.5:1; UI elements ≥ 3:1)

Assertion-Evidence writing tips

references/STYLE-GUIDE.md

Layout & Style (neutral theme, supports brand replacement):

Canvas: 16:9; safe margins ≥ 48px; grid column spacing 24px

Fonts: Chinese (Source Han Sans/PingFang/Hiragino Sans), English (Inter/Calibri)

Font sizes: Heading 34-40, Subheading 24-28, Body 18-22, Footer 14-16

Line spacing: Heading 1.1, Body 1.3; bullet spacing ≥ 8px

Color palette (AA compliant): Dark ink #1F2937 / Background #FFFFFF / Accent #2563EB / Emphasis #DC2626

Components: unified 6-8px border radius; charts and images with 8px padding

Images: add brief alt descriptions for screen readers

Page density: ≤ 70 words per slide (excluding captions)

references/RUBRIC.md

PPT Quality Scoring Rubric (100 points; ≥ 75 to deliver):

Each item scored 0-10:

Goal Clarity: Audience, objective, CTA well-defined

Story Structure: Pyramid structure complete, hierarchy clear

Slide Assertions: Headings are "assertion sentences" supported by evidence

Evidence Quality: Data/cases/citations sufficient, credible, consistent calibration

Chart Fit: Correct selection, complete labeling, readable

Visual & Accessibility: Contrast, font size, white space, color compliance

Coherence & Transitions: Natural chapter and page transitions

Speakability: 45-60 sec per slide, natural language

Deliverables Complete: slides.md / notes.md / refs.md / (optional) assets/*.png

Robustness: Gaps explicitly marked, fallback plan & next steps provided

Self-evaluation process: Run CHECKLIST first, then score each item and write top 3 low-scoring items + improvement actions. If total < 75, apply actions and re-score (max 2 iterations).

references/CHECKLIST.md

Pre-flight checklist for final quality assurance before delivery.

references/EXAMPLES.md

Two Usage Examples:

Example A: Ultra-simple topic ("coffee") → trigger minimal questionnaire, generate 12-page framework with placeholder charts

Example B: Small business monthly review with attached CSV → parse data, select charts per VIS-GUIDE, call chartkit.py, refine 1-2 iterations if score < 75

scripts/chartkit.py

Minimal chart renderer for ppt-creator.

Usage:

python scripts/chartkit.py \

  --data path/to/data.csv \

  --type line \

  --x date \

  --y sales profit \

  --out output/assets \

  --filename kpi_trend.png \

  --title "Monthly KPIs"

Notes:

Requires: pandas, matplotlib

Fallback: If packages unavailable, print instruction message and exit(0)

Uses matplotlib defaults for readability (no hard-coded brand colors)

Advanced Tips

This skill complements (not conflicts with) Anthropic's built-in PowerPoint generation capabilities. Use this skill to produce "high-quality structure & content," then optionally invoke system capabilities to export the final PPTX file.

For complex data analysis needs, combine with other skills (e.g., data analysis, charting) before invoking ppt-creator.

The skill is designed to be forgiving: missing information triggers safe defaults rather than blocking progress.

ppt-creator