
from presentation-skills124
Assist agents in producing polished, editable PowerPoint decks end-to-end: planning, slide archetypes, asset slotting (charts, diagrams, images), preview export
This skill provides a deck-level workflow and quality gate system for creating high-quality, editable PPTX deliverables. It guides the agent through establishing a deck_contract (audience, template constraints, visual and typography tokens), deriving slide contracts, registering asset_slots for charts/figures/diagrams/images, assembling native PPTX, and running validation and preview checks. The skill includes scripts and references to automate workspace initialization, template auditing, preview export, and validation steps.
Use this skill when the user needs a maintainable, shareable PowerPoint deck (roadshow, research report, technical proposal, training material) — especially when the output must be editable PPTX, include charts/diagrams that remain editable where possible, and pass visual/structure validation before handoff. It's suited to formal external decks, template-bound builds, and cases requiring preview exports and human review checkpoints.
Compatible with agent tooling that can run Python scripts and manage file-based workflows (Codex CLI, Claude Code-style agents, Copilot/Copilot-like runners). The repo targets Codex CLI workflows but the patterns are broadly applicable to agents that can execute the included scripts and assemble PPTX files.
A comprehensive PPT deck creation and collaboration skill with 15 scripts covering workspace init, template auditing, quality checks, preview export, image generation via OpenAI API, and slide spec derivation. The SKILL.md is extremely detailed (12K+ chars, Chinese-language) with a well-structured workflow. Most scripts failed in test due to missing python-pptx and other dependencies; only python_figure_helpers.py (which lazy-loads matplotlib/seaborn) succeeded. The skill uses proper argparse, error handling, and .env file loading for API keys.
python-pptxfitz (PyMuPDF)pandasseabornmatplotlibopenaiPillowNo hardcoded credentials found. API keys loaded from env vars or .env files (proper practice). urllib.request.urlopen used for downloading generated images and SVG icons — standard pattern but worth noting. subprocess.run used only for qlmanage (macOS SVG conversion) — no shell injection risk as it uses list form. The skill is a legitimate, well-architected PPT production pipeline. Security deductions: -8 for urllib.urlopen on external URLs without certificate pinning, -5 for subprocess usage (macOS-specific, list form), -5 for OpenAI API key passed via CLI arg (--api-key) which could leak to process lists.