AIPPT

Installation

npx skills add bahayonghang/AIPPT

AIPPT is a Claude Code skill for brand-new deck projects. It is contract-first: research, argument, layout, style, and delivery are staged and validated before handoff.

Use it for full new-deck work, including staged checkpoints like outline_only and spec_only when they belong to a larger deck build.

Do not use AIPPT for:

• casual outline-only brainstorming with no contract path
• critique-only or review-only requests
• single-slide work
• export-only or conversion-only work
• template tweaks
• editing existing .pptx/.ppt/.key/Google Slides files

Current skill surface

AIPPT now routes through two layers:

• generic AIPPT: the default staged workflow
• scene packs: narrow subskills for recurring deck types
• golden-path.md: the shortest default operator path for standard new-deck requests
• ./.aippt/EXTEND.json and ~/.aippt/EXTEND.json: project/user defaults for scene, style, delivery mode, language, strict review, and style dimensions
• dual-layer style + layout system: preset/style dimensions plus layout archetype/final geometry
• scriptable delivery surface: prompt bundle, SVG validation, static preview, and scene-pack scaffolding

Built-in scene packs:

• company-intro
• investor-pitch
• board-briefing
• policy-briefing
• teaching-deck
• thesis-defense

The root skill stays the contract layer. It does not become a slide editor runtime.

Golden path

The shortest default operator path is:

1. route + intake
2. brand_profile + brief_summary
3. research_dossier
4. outline hard stop
5. slide_spec
6. page_plan
7. style_profile
8. delivery + validation

Hard gate: the first outline must keep approved=false.

Scene packs only refine defaults; they never bypass gates.

Output model

Default workspace:

output/
├── briefing/
├── specs/
├── prompts/
├── svg/
├── preview/
└── project.json

Delivery modes:

• outline_only
• spec_only
• prompt_bundle_only
• svg_pages
• brand_ready_assets

Key commands / script entrypoints

• cd docs && npm run aippt:list-catalog
• cd docs && npm run aippt:init-workspace -- --output-dir ../output --scene-id investor-pitch
• node skills/aippt/scripts/read-preferences.mjs
• cd docs && npm run aippt:create-scene-pack -- --id customer-story --label "Customer Story" --description "Use first-party case proof and before/after narrative."
• cd docs && npm run aippt:build-prompts -- --slide-spec ../output/specs/slide-spec.json --page-plan ../output/specs/page-plan.json --brand-profile ../output/briefing/brand-profile.md --style-profile ../output/specs/style-profile.json --scene-pack investor-pitch --output-dir ../output/prompts
• cd docs && npm run aippt:validate-artifacts -- --outline ../output/specs/outline.json --slide-spec ../output/specs/slide-spec.json --page-plan ../output/specs/page-plan.json --style-profile ../output/specs/style-profile.json --delivery-manifest ../output/prompts/delivery-manifest.json --scene-pack investor-pitch
• cd docs && npm run aippt:validate-svg -- --input ../output/svg --page-plan ../output/specs/page-plan.json --manifest ../output/prompts/delivery-manifest.json
• cd docs && npm run aippt:build-preview -- --svg-dir ../output/svg --output ../output/preview/index.html --manifest ../output/prompts/delivery-manifest.json

Canonical maps

Use these as source of truth:

• skills/aippt/references/resource-registry.md
• skills/aippt/references/golden-path.md
• skills/aippt/references/scenes/scene-catalog.json

Evaluation

Regression sources:

• skills/aippt/evals/evals.json
• skills/aippt/evals/trigger-evals.json
• skills/aippt/references/eval-prompts.md
• skills/aippt/evals/scene-stubs/README.md + generated stubs as scaffold coverage to promote into the main suites

Docs

• Chinese README: README_CN.md
• VitePress docs: docs/

cd docs
npm install
npm run docs:dev