Skip to content

nexu-io/open-design / plugins/community/humanize-ppt

humanize-ppt

Verified

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 with per-page visual-enhancement decisions (image / SVG diagram / video), hands a production brief to a downstream renderer (HTML-PPT skills or native PPTX), then runs a capped 3-round presentation checkup (演讲体检) on the rendered deck. It never renders slides itself. Use before generating PPT/HTML slides from raw material, and after rendering when the user says things like "给这份 deck 做演讲体检" or "PPT 渲染质检". If all you want is one beautiful template page with no outline and no checkup, a rendering skill alone is enough.

FileKindSize
LICENSEother219 B
SKILL.mdskill md15.6 KB
SPEC.mdmarkdown13.1 KB
adapters/guizang-bridge-notes.mdmarkdown153 B
adapters/ppt-master-bridge-notes.mdmarkdown3.3 KB
adapters/presenter-adapter-spec.mdmarkdown458 B
adapters/zara-bridge-notes.mdmarkdown175 B
contracts/asset-manifest.template.mdmarkdown75 B
contracts/ast-outline.schema.mdmarkdown211 B
contracts/deck-brief.template.mdmarkdown114 B
contracts/router-plan.schema.jsonconfig1.2 KB
contracts/run-manifest.schema.jsonconfig608 B
contracts/slide-plan.schema.jsonconfig4.4 KB
contracts/speaker-intent.template.mdmarkdown61 B
contracts/video-slots.schema.jsonconfig528 B
docs/index.mdmarkdown9.2 KB
docs/local-demo.mdmarkdown3.2 KB
examples/01-ai-tool-update/source.mdmarkdown940 B
open-design.jsonconfig1.6 KB
references/agent-teams-public-preview.mdmarkdown2.9 KB
references/beautiful-preview-first-adapter.mdmarkdown2.7 KB
references/guizang-material-qa.mdmarkdown5.2 KB
references/guizang-presenter-deploy.mdmarkdown1.9 KB
references/guizang-production-brief-orchestrator.mdmarkdown5.9 KB
references/humanize-ppt-public-writing.mdmarkdown2.5 KB
references/presenter-export-adapter.mdmarkdown2.5 KB
references/preview-outline-spec.mdmarkdown7.3 KB
references/qa-failure-modes.en.mdmarkdown18.7 KB
references/qa-failure-modes.mdmarkdown16.6 KB
references/renderer-guidance.mdmarkdown2.1 KB
references/renderer-verification.mdmarkdown2.9 KB
references/selected-template-full-deck-adapter.mdmarkdown2.5 KB
references/style-gallery-spec.mdmarkdown5.1 KB
references/workbuddy-team-packaging-and-video-materials.mdmarkdown5.2 KB
registry/renderer_registry.jsonconfig11.4 KB
scripts/humanize_ppt.pyscript349 B
scripts/humanize_ppt_v1.pyexecutablescript11.0 KB
scripts/humanize_ppt_v2.pyexecutablescript174.5 KB
scripts/humanize_ppt_v3.pyscript331 B
scripts/humanize_ppt_v4.pyscript350 B
scripts/humanize_ppt_v5.pyscript354 B
scripts/pptx_qa.pyscript14.9 KB
scripts/preview_outline_html.pyscript9.6 KB
scripts/smoke_check.pyscript11.4 KB

references/style-gallery-spec.md

View on GitHub
# v0.9 Spec: `--style-gallery` Flag (the cover-style gate)

Date: 2026-06-17
Status: Implemented in v0.9 (`feat/v0.9-next`).

v1.1 renderer exception: PPT Master already owns a mandatory three-stage Confirm UI with native `visual_style` preview SVGs. `--renderer ppt-master --style-gallery` therefore delegates to that surface; it does not create four speculative PPT Master projects.

## Motivation

Humanize PPT is *for the presentation* and *compatible with downstream HTML PPT renderers* — and it **never renders**. By v0.8 the flow was: AST slice → (optional `--preview-outline` review) → production brief → downstream renders. The human picks the renderer and (for guizang) a single style/theme up front, blind. They commit to one look before seeing a single pixel.

The 9-style showcase on "10 分钟入门 Agent 7 概念" showed the cost: people want to *see covers side by side* before committing. One static thumbnail per style is the cheapest possible preview — a cover is the highest-signal page, and rendering one page is cheap for the downstream skill.

The architecture decision (拍板): **the gallery shows ≥4 covers that the downstream skill really renders.** Humanize stays in its lane — it emits the spec and the commands; it does not render the covers itself.

## Design

### New flags

```
--style-gallery            (boolean, default off) — the gate before the outline.
--gallery-count N          (int, default 4, minimum 4) — capped at the candidates defined for the renderer.
```

`--style-gallery` is checked **before** `--preview-outline` in `main()`: you pick a style, *then* preview the outline. Passing both stops at the gallery.

### What it writes (then stops)

For the resolved primary renderer (`choose_routes`), Humanize takes the first
`N` candidates from `STYLE_GALLERY_CANDIDATES[renderer]` and writes:

1. `commands/style-gallery/<id>.md` — a **cover-only** render command per
   candidate. It tells the downstream skill to render **only S01** in that
   style and write `outputs/style-gallery/<id>/cover.{html,png}`. Not a full deck.
2. `style_gallery.html` — a **zero-dependency single-file** picker that stitches
   the candidate covers (relative-path `<iframe>`s) with each candidate's label,
   description, and the exact re-injection command. Honest about pending state:
   a not-yet-rendered cover shows the frame backdrop plus an always-visible
   caption — no faked thumbnail (宁空不摆拍).
3. `style_gallery_plan.json` — machine-readable: per-candidate `id`, `label`,
   `description`, `cli`, `command_file`, `cover_html`, `cover_png`,
   `reinjection_command`.

Then it prints `{"ok": true, "stopped_at": "style-gallery", ...}` and returns 0.
No outline, no brief, no QA.

For `ppt-master`, the renderer-native branch writes instead:

1. `commands/style-gallery/ppt-master-confirm-ui.md` — re-injection command plus the instruction to use PPT Master's Stage 1 direction page.
2. `style_gallery_plan.json` with `mode: downstream-confirm-ui`, `picker: null`, and `candidate_source` pointing to PPT Master's own visual-style catalog.

It returns `stopped_at: ppt-master-style-gate`. No fake `style_gallery.html`, cover image, or candidate deck is emitted.

### The candidates

Module-level `STYLE_GALLERY_CANDIDATES`, keyed by renderer, ≥4 each:

- **guizang** — spans both tracks so the four covers are visually distinct:
  Style A `ink-classic` (the known-good baseline), Style A `kraft-paper`,
  Style A `indigo-porcelain`, Style B Swiss `ikb`.
- **frontend-slides** — 4 style directions (editorial-serif, techno-grid,
  soft-gradient, mono-contrast).
- **beautiful-html-templates** — 4 template slots filled from the skill's
  native template library (placeholder slugs; downstream picks real templates).

Each candidate carries `cli`: the exact renderer/style args grafted onto the
next run. After picking a cover, the human runs the printed
`reinjection_command`, which resumes the normal outline → brief flow with the
chosen style. (Append `--preview-outline` for the review checkpoint.)

### WebGL static-screenshot trap (ties to QA failure modes)

Guizang Style A covers use a WebGL hero canvas. A static PNG screenshot
captures the canvas **before it paints → blank cover** (实证: a Style A
`ink-classic` cover PNG came back 14KB, an empty page). So each Style A
candidate command warns:

- treat `cover.html` (the live page) as the source of truth; `cover.png` is a
  thumbnail only,
- wait for the canvas's first frame before screenshotting (delay ≥1.5s),
- a `cover.png` under 20KB is a failed capture, not an empty cover.

The picker footer repeats this so a blank frame is read correctly.

## Boundary (unchanged)

Humanize decides *what* (which candidate styles) and *where* (the output
paths). The downstream skill produces the cover files. Humanize never opens a
renderer template and never post-processes rendered HTML. The picker is a
Humanize-owned working draft (like `preview_outline_html.py`), not a deck.

## Reuse

`run_style_gallery_mode` reuses `read_source` / `detect_language` /
`build_slide_plan` / `choose_routes` / `now_iso` — the same primitives as the
brief and preview-outline modes. No new parsing path.
canvas-designanthropics/skillsCreate beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
docxanthropics/skillsUse this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pdfanthropics/skillsUse this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
pptxanthropics/skillsUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates (.potx), layouts, speaker notes, or comments. Trigger whenever the user mentions "deck," "slides," "presentation," or references a .pptx or .potx filename, regardless of what they plan to do with the content afterward. If a .pptx or .potx file needs to be opened, created, or touched, use this skill.Stars: 198kUpdated Claude CodeCodexVerifiedInstall