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 before rendering slides.
Positioning
Humanize PPT is an Outline Director and Agent Teams Orchestrator, not a slide renderer.
It should run before downstream PPT / HTML slide skills. Its job is to produce a clean AST-based production brief so renderers do not ingest raw noisy material directly.
In Agent Teams mode, the main Humanize PPT Agent loads this skill and controls specialist agents:
- Guizang Agent for Chinese stable rendering.
- Zara Agent for style exploration, HTML production, and deploy.
- HyperFrames Agent for video slots.
- Presenter Agent for presenter mode after the deck is finalized.
- QA Agent for content, visual, path, and delivery checks.
AST theory
AST means Audience-State-Transfer.
- Audience: who is listening, what they know, what they resist, and why they would keep listening.
- State: the audience state before and after the deck, plus the core tension that blocks the transition.
- Transfer: the slide-by-slide path that moves the audience from initial state to desired state.
Core sentence:
PPT is not an information container. PPT is an audience state-transfer artifact.
Required output contract
For every Humanize PPT run, produce:
deck_brief.md — audience, goal, tension, success criteria.ast_outline.md — AST map and narrative arc.slide_plan.json — slide-by-slide plan.speaker_intent.md — what the speaker should do on each slide.asset_manifest.md — screenshots, charts, images, video needs.video_slots.json — optional HyperFrames / video insertion plan.
Recommended OPC workflow
O — Outline Director
Humanize PPT: raw material → AST outline + production brief
P — Presentation Production
guizang path: Chinese stable HTML PPT
Zara path: style exploration and HTML production
C — Complete / Control
HyperFrames video adapter
Presenter Adapter shell
Deploy / export adapter
QA checklist
Rules
- Do not let slide renderers consume raw material directly when Humanize PPT can first produce the AST contract.
- Keep presenter mode as a post-processing adapter, not a style.
- Separate deployment from presenter mode.
- Absorb AI-writing cleanup principles from humanizer tools, but do not reduce Humanize PPT to text polishing.
- Prefer a small verified workflow over a broad unverified promise.
- For public Skill releases, create/push the repo, install from GitHub locally, run one safe full sample, verify style exploration + presenter mode + deploy URL, and only then polish README details.
- For Agent Teams development, emit
router_plan.json, run_manifest.json, bounded commands/*.md, and separate outputs/<agent>/ directories before wiring real downstream Skills.
Operational references
references/agent-teams-public-preview.md — Agent Teams architecture, specialist-agent command protocol, public preview release loop, and README split convention.
Local demo
If this repository is installed locally, run:
python3 scripts/humanize_ppt_v1.py \
--source examples/01-ai-tool-update/source.md \
--out .humanize-ppt-runs/ai-tool-update \
--title "AI 工具更新，不只是功能清单"
1---2name: humanize-ppt3description: AST-based outline director for human-centered AI presentation workflows. Use before generating PPT/HTML slides from raw material.4license: MIT5---6
7# Humanize PPT
8
9Use this skill when a user wants to turn raw material, notes, voice transcripts, documents, links, or old PPTs into a presentation-ready outline before rendering slides.
10
11## Positioning
12
13Humanize PPT is an **Outline Director** and **Agent Teams Orchestrator**, not a slide renderer.
14
15It should run before downstream PPT / HTML slide skills. Its job is to produce a clean AST-based production brief so renderers do not ingest raw noisy material directly.
16
17In Agent Teams mode, the main Humanize PPT Agent loads this skill and controls specialist agents:
18
19- Guizang Agent for Chinese stable rendering.
20- Zara Agent for style exploration, HTML production, and deploy.
21- HyperFrames Agent for video slots.
22- Presenter Agent for presenter mode after the deck is finalized.
23- QA Agent for content, visual, path, and delivery checks.
24
25## AST theory
26
27AST means **Audience-State-Transfer**.
28
29- **Audience**: who is listening, what they know, what they resist, and why they would keep listening.
30- **State**: the audience state before and after the deck, plus the core tension that blocks the transition.
31- **Transfer**: the slide-by-slide path that moves the audience from initial state to desired state.
32
33Core sentence:
34
35> PPT is not an information container. PPT is an audience state-transfer artifact.
36
37## Required output contract
38
39For every Humanize PPT run, produce:
40
411. `deck_brief.md` — audience, goal, tension, success criteria.
422. `ast_outline.md` — AST map and narrative arc.
433. `slide_plan.json` — slide-by-slide plan.
444. `speaker_intent.md` — what the speaker should do on each slide.
455. `asset_manifest.md` — screenshots, charts, images, video needs.
466. `video_slots.json` — optional HyperFrames / video insertion plan.
47
48## Recommended OPC workflow
49
50```text
51O — Outline Director
52 Humanize PPT: raw material → AST outline + production brief
53
54P — Presentation Production
55 guizang path: Chinese stable HTML PPT
56 Zara path: style exploration and HTML production
57
58C — Complete / Control
59 HyperFrames video adapter
60 Presenter Adapter shell
61 Deploy / export adapter
62 QA checklist
63```
64
65## Rules
66
671. Do not let slide renderers consume raw material directly when Humanize PPT can first produce the AST contract.
682. Keep presenter mode as a post-processing adapter, not a style.
693. Separate deployment from presenter mode.
704. Absorb AI-writing cleanup principles from humanizer tools, but do not reduce Humanize PPT to text polishing.
715. Prefer a small verified workflow over a broad unverified promise.
726. For public Skill releases, create/push the repo, install from GitHub locally, run one safe full sample, verify style exploration + presenter mode + deploy URL, and only then polish README details.
737. For Agent Teams development, emit `router_plan.json`, `run_manifest.json`, bounded `commands/*.md`, and separate `outputs/<agent>/` directories before wiring real downstream Skills.
74
75## Operational references
76
77- `references/agent-teams-public-preview.md` — Agent Teams architecture, specialist-agent command protocol, public preview release loop, and README split convention.
78
79## Local demo
80
81If this repository is installed locally, run:
82
83```bash
84python3 scripts/humanize_ppt_v1.py \
85 --source examples/01-ai-tool-update/source.md \
86 --out .humanize-ppt-runs/ai-tool-update \
87 --title "AI 工具更新，不只是功能清单"
88```
ahang1598