| name | humanize-ppt |
| description | A presentation system for agent-made PPTs — born for the talk, not just the template. It turns raw material into an AST (audience-state-transfer) outline where every page turn moves the audience forward, decides per-page visual enhancement (real images via baoyu-image-gen on the local Codex CLI, SVG diagrams, Remotion video), then runs a capped 3-round presentation checkup (演讲体检) on the rendered HTML, comparing each page against its outline page and writing fix prompts. The beautiful deck is rendered natively by a downstream template skill (broadly compatible with any HTML-PPT skill; verified: guizang-ppt-skill for Chinese, frontend-slides / beautiful-html-templates for English) — Humanize orchestrates the presentation, the template skill paints the slides. Use before generating PPT/HTML slides from raw material, and after rendering when the user says things like "帮我盯一下渲染出来的 PPT 有没有翻车", "PPT 渲染质检", "给这份 deck 做演讲体检", or "告诉我哪几页只能看不能讲". If all you want is one beautiful template page with no outline and no checkup, a rendering skill alone is enough; bring in Humanize when you need the talk's narrative before rendering and the checkup after. |
| version | 1.0.0 |
| author | LearnPrompt |
| license | MIT |
| requires-skills | {"guizang-ppt-skill":"Required when the deck language is Chinese. The brief writer references ~/.agents/skills/guizang-ppt-skill/SKILL.md; without it the next agent cannot render. v0.6.5 brief writer emits a stderr warning if the skill is not detected.","frontend-slides":"Recommended when the deck language is English. Same hand-off pattern as guizang. v1.0 support_level: full — brief exit works, full presentation checkup ran on a real deck on 2026-06-17 (docs/showcase/v0.9-frontend-slides/), and 5 renderer-specific failure-mode rules added (horizontal overflow, low contrast, hyphenation noise, font contract, image alt).","beautiful-html-templates":"English alternative. Same hand-off pattern. v1.0 support_level: full — brief exit works, full presentation checkup ran on a real Neo-Grid deck on 2026-06-13 (docs/showcase/hermes-agent-mastery/en/), same 5 renderer-specific failure-mode rules as frontend-slides.","remotion-video-production":"Recommended (main) for the video media slot. Orchestrates the whole Remotion video pipeline — renders the real mp4 (deterministic loop, no narration) to the slot's asset_path. Verified in v0.9 (docs/showcase/v0.9-visual-enhancement/).","remotion-best-practices":"Pair with remotion-video-production while writing Remotion code — avoids unstable patterns (misused CSS / Tailwind animation, wrong asset paths).","remotion-video-toolkit":"Add only for complex video work — captions, charts, 3D, batch templates, automated render pipelines. Source: github.com/shreefentsar/remotion-video-toolkit.","baoyu-image-gen":"Recommended for the image media slot. Drives the local Codex CLI (--provider codex-cli) using the logged-in Codex/ChatGPT subscription — no OPENAI_API_KEY needed. Generates the real hero/concept image to the slot's asset_path. Source: github.com/JimLiu/baoyu-skills. Verified in v0.9."} |
| metadata | {"category":"content","tags":["presentation","ppt","html-slides","humanizer","ast","workflow","brief-orchestrator","hv-analysis","9-styles"]} |
Humanize PPT
Use this skill when a user wants to turn raw material, notes, voice transcripts, documents, links, or old PPTs into a presentation-ready outline and per-page media decisions before delegating rendering to a downstream skill.
Positioning
Humanize PPT is a presentation system, born for the talk: an Outline Director (AST audience-state-transfer — every page turn moves the audience forward), a Per-Page Visual-Enhancement Director (real image / SVG diagram / Remotion video), a Production Brief Orchestrator, a Presentation Checkup Runner (演讲体检; formerly the QA loop, CLI flag still --qa-from), and a Presenter-Mode hand-off. The motivation: HTML-PPT template skills are great at concept display but blow a simple idea into a dozen pretty pages, while a real 90-minute talk is ~30 — the pretty shell outruns the content density. Humanize closes that gap: it keeps the beauty (rendered natively by the downstream template skill) and makes it presentable — a line you can stand up and deliver. Downstream template skills own "renders beautifully"; Humanize owns "it's a talk, and someone checked it."
The presentation checkup in one sentence: it does not grade beauty, it grades the outline. It compares every rendered page against its outline page, pulls out the pages that can only be looked at but not spoken from, and keeps going until every page is one the speaker can stand up and present. A failed page, in plain words: a page that holds only a few words and never finishes its point, or a page that fails the audience state transfer it promised (the listener walks out of that page in the same state they walked in). Such a page should not exist; the checkup pulls it out and generates fix instructions.
Humanize is broadly compatible with any downstream skill that can output an HTML PPT: the brief is plain markdown + JSON, anything can read it. The verified, stable recommendations are: Chinese → guizang-ppt-skill; English → frontend-slides / beautiful-html-templates. Other downstreams are hot-pluggable; support levels live in registry/renderer_registry.json and are updated only on real results.
It runs before downstream PPT / HTML slide skills and around the post-render presentation checkup. It owns the AST contract, the per-page media decision (does this page need a photo, a system diagram, a 10-second process clip, nothing?), the production brief that the next agent consumes, and the checkup pass on the rendered HTML. It does not own the rendered HTML itself.
There are two human review gates before rendering, both Humanize-owned zero-dependency working drafts (never decks): the ( / , the audience state-transfer map) and, since , the (). The style gallery is the cover-style gate: before the outline, Humanize emits ≥4 cover candidates for the renderer, writes one cover-only render command per candidate for the downstream skill to render, and stitches to pick from. Humanize emits only the spec and the commands; the covers are rendered downstream. Picking a cover yields a re-injection command that carries the chosen style into the normal flow. See .
