Codex PPT Skill

What Codex PPT Skill does

Codex PPT Skill plans an outline and a visual style, generates each slide as a full 16:9 image with gpt-image-2, and then assembles the images into a .pptx file with a local script. Input can be articles, reports, papers, course notes, Markdown, outlines, PDF or Word files. The workflow is staged: it asks you to confirm outline.md, per-slide points, the style direction and a sample slide before generating the full deck.

It ships 12 built-in style references, such as clean professional, research defence, hand-drawn technical explainer and McKinsey-style, and can imitate a style from an image, PDF or PPTX you upload. Finished styles can be saved to a personal library at ~/.codex-ppt-skill/references/, which survives skill updates. After the sample is approved, subagents can generate one slide each and self-check text clarity and consistency. It also writes a speech.md and puts speaker notes into the PPTX.

Key features

• Slides generated as 16:9 images and assembled into .pptx by a local script
• Staged workflow: outline, style, sample slide, then the full deck
• 12 built-in style references and custom style replication from uploaded files
• Personal style library stored outside the skill install directory
• Optional one-subagent-per-slide generation with self-checks
• Generates speech.md and writes speaker notes into the PPTX

When to use it

• Turn a research paper into a defence-style deck with original figures placed on chosen slides
• Convert course notes or a report into a presentation draft
• Reuse a favourite slide style from a PDF or PPTX on new content

Who it is for: People who want AI-generated presentations from existing documents and use Codex or another SKILL.md-compatible agent.

How it fits with Hermes Agent

Supports Hermes Agent among other SKILL.md agents such as Codex, Claude Code and OpenClaw; outside Codex it needs an external image-generation setup.

Requirements: Outside Codex: gpt-image-2.5-flare, a third-party image generation API, or an OpenAI-compatible image endpoint

Note: Slides are exported as images, so individual elements are not editable, and agents other than Codex need a separate image-generation API.

FAQ