Skip to content

AI turns documents or topics into real, native PowerPoint decks—with native shapes, transitions and animations, data-backed charts and tables on demand, audio narration from speaker notes, and support for your own .pptx templates. · by Hugo He

refactor(ai-image): drop defensive defaults and scenario whitelists from image rules

hugohe32344e7085294
diff --git a/skills/ppt-master/references/image-generator.md b/skills/ppt-master/references/image-generator.md
index b65c4327..11bda6b2 100644
--- a/skills/ppt-master/references/image-generator.md
+++ b/skills/ppt-master/references/image-generator.md
@@ -10,24 +10,30 @@ Role definition for the **AI image generation path**: convert each `Acquire Via:
## 1. Core Principle — Maximize AI Image Capability in Service of the Deck
-**Hard rule**: AI images exist to serve the deck's communication goal. Default boundaries (`page_role: local`, `text_policy: none`) are starting points, not ceilings. Promote to `hero_page` or `embedded` when doing so makes the page work better — provided the deck's rendering and palette lock hold.
+AI images exist to serve the deck's communication goal. Pick whatever combination of `page_role` and `text_policy` makes the page work best.
-**Two page roles**:
+**Two page roles** (orthogonal to type):
-| `page_role` | Image's job | SVG above |
-|---|---|---|
-| `local` (default) | A region block on an SVG page; carries visual, not content. Reserve 12-20% inner padding so content survives container cropping | Carries headline / body / labels |
-| `hero_page` | Image is the page's main voice — cover, chapter divider, mood transition, data hero, closing quote | Minimal overlay only (page number, brand bar, chapter mark); may be empty when the image already carries equivalent design elements |
+| `page_role` | Use |
+|---|---|
+| `local` | Image occupies a region of an SVG page (left half, right column, hero band, accent corner). Composition is the AI's call — fill the region as the page design wants |
+| `hero_page` | Image is the page's main voice — cover, chapter divider, mood transition, single-number hero, closing quote. SVG above may be minimal or empty |
-**Two text roles inside the image** (orthogonal to `page_role`):
+**Two text policies** (orthogonal to page_role):
-| Role | Example | Prompt treatment |
-|---|---|---|
-| Decorative | "GROWTH" as background lettering, retro stamp, scattered alphabet texture | Describe artistic direction; spelling is not critical; let the model improvise |
-| Design-title | Cover main title, chapter heading | Name the font family (must echo deck's SVG typography); content must be accurate |
-| Content (never inside the image) | Body copy, data points, bullet lists, long quotes | Route to SVG |
+| `text_policy` | Use |
+|---|---|
+| `none` | No text inside the image |
+| `embedded` | Image contains text as part of the artwork — decorative lettering, designed title, hand-lettered keywords, infographic labels, anything the page needs |
+
+**Hard rule — only what's actually hard**:
+
+- Same `deck_rendering` + same `deck_palette` for every image in the deck
+- HEX codes and color names are rendering guidance — never visible text in the image
+- Long body copy / data points / bulleted lists / long quotes stay in SVG (improving them later means regenerating the image, which is expensive)
+- Prompts are one coherent prose paragraph, not tag soup (a model-output reality, not an aesthetic choice)
-**Hard rule — deck-wide coherence still holds**: every AI image shares the deck's locked rendering and palette regardless of `page_role` or `text_policy`. `hero_page` is not permission to switch rendering for one slide.
+Everything else is the AI's judgment per page. No mandated padding, no type-locked text_policy, no scenario whitelists for hero_page.
---
@@ -108,8 +114,8 @@ Then `read_file` the **single resolved** rendering file and the **single resolve
For each `Acquire Via: ai` row in `design_spec.md §VIII`:
1. **Determine type** by matching the row's `Purpose` against types `_index.md` auto-selection table (cover background → `background`; product launch hero → `hero`; methodology visualization → `framework`; etc.) The narrative-shorthand `Type` column in §VIII (Background/Photography/Illustration/Diagram/Decorative) is a hint, not the type's final value — `Purpose` is authoritative for picking among the 9 internal-composition types.
-2. **Determine `text_policy`** — Strategist's value (when present) wins. Otherwise judge per-image: default `none` (image carries no text, SVG overlays labels); upgrade to `embedded` when the page would communicate better with in-image text — covers / chapter dividers with a designed title, typography pieces, infographics with internal labels, sketch-notes with hand-lettered words, posters with decorative lettering. Do not put body copy, data points, or long quotes inside the image.
-3. **Determine `page_role`** — Strategist's value (when present) wins. Otherwise default `local`. Upgrade to `hero_page` when the page's job is "image speaks louder than text" — cover, chapter divider, mood transition, data hero, closing quote.
+2. **Determine `text_policy`** — Strategist's value wins when set. Otherwise pick `none` or `embedded` based on whether in-image text serves the page. Long body / data / lists stay in SVG.
+3. **Determine `page_role`** — Strategist's value wins when set. Otherwise pick `local` or `hero_page` based on whether the image carries the page or sits inside one.
4. `read_file references/image-type-templates/<type>.md` (only if not already read — types are commonly reused across images in one deck)
5. **Assemble the prompt** by combining:
- The rendering's style paragraph (from Step 2)
@@ -136,11 +142,11 @@ Every assembled prompt follows this paragraph structure. **Write prose, not tag
[Palette behavior — apply the chosen palette's proportion + role rules to the deck's HEX values, e.g. "primary #1E3A5F dominates as the main shape, secondary #F8F9FA provides 60% breathing space, accent #D4AF37 appears in one or two emphasis points only"].
[Type-specific composition — from the chosen type file, e.g. "central hub node with four radiating satellite nodes connected by clean lines"].
[Image-specific subject — translated from the row's Reference intent into concrete visual nouns].
-[Container reminder — "composed as a local visual block with 15% padding on all sides, designed to be embedded within a slide region of roughly {W}x{H}px, leaves negative space around edges"].
+[Container note — "composed as a {W}x{H}px image for {page_role} use"; add composition cues only when the page actually needs them (e.g. "leave the lower band relatively calm — SVG title overlays it")].
[Hard rules — see §5].
```
-**Word budget**: 150-250 words for `text_policy: none`; 180-300 words for `text_policy: embedded`. `hero_page` items skew toward the upper end of their bracket — the image carries more design weight.
+**Word budget**: 150-300 words. Embedded-text prompts skew longer; pure background prompts can be shorter.
**Forbidden — tag-soup prompts**:
@@ -174,31 +180,25 @@ Exception: when the chosen rendering is `corporate-photo`, photorealism is inten
### 5.3 Text policy — none vs embedded
-| `text_policy` | What the image contains | Append to prompt |
-|---|---|---|
-| `none` (default) | Zero text, letters, numbers, labels | "NO text of any kind. No letters, numbers, signs, watermarks, labels, or written symbols anywhere in the image. Clean negative space ready for SVG text overlay." |
-| `embedded` | Text is part of the artwork — decorative lettering, designed title, hand-lettered keywords, infographic labels | See "Embedded prompt construction" below |
+| `text_policy` | Prompt cue |
+|---|---|
+| `none` | "NO text of any kind anywhere in the image — no letters, numbers, signs, watermarks, labels, or written symbols." |
+| `embedded` | Describe the text directly inside the visual scene: the word(s), how they're rendered (decorative lettering / designed title / hand-lettered keyword), and the artistic treatment. Examples below. |
-**Embedded prompt construction**: name the text role explicitly so the model knows how to treat it.
+**Prompt phrasing examples for embedded text** (not an exhaustive list):
-| Text role | Prompt phrasing |
-|---|---|
-| Decorative | "decorative '{WORD}' lettering as a background element, {style describing artistic treatment — 3D extruded retro, neon glow, hand-painted brush, scattered alphabet texture, etc.}" — let the model improvise; spelling is not critical |
-| Design-title | "main title text '{exact words}', rendered in {font family — clean sans-serif / serif editorial / hand-written / geometric display} echoing the deck's body typography" — content must be accurate |
-| Hand-lettered keyword set | "1-5 short hand-lettered keywords in English: '{w1}', '{w2}', '{w3}' — woven into the composition as sketch-notes annotations" |
+- Decorative: "large 'GROWTH' lettering as a background element, 3D extruded retro chrome style"
+- Designed title: "main title 'Q3 STRATEGY' typeset in clean geometric sans-serif, centered"
+- Hand-lettered set: "small hand-lettered annotations 'fast', 'cheap', 'good' woven into the sketch"
-**Font family must echo the deck**: when Strategist locked `deck_typography` in `spec_lock.md`, use that family. Otherwise pick a family compatible with the chosen rendering (vector-illustration / flat → clean sans-serif; editorial → serif; sketch-notes / ink-notes → hand-written; 3d-isometric / digital-dashboard → geometric display).
+Font choice and visual treatment are part of the page's design — pick what fits. The deck's `image_rendering` already implies a typographic mood (sketch-notes → hand-drawn lettering; editorial → serif; 3d-isometric → geometric display), but no specific family is required.
-**CJK warning**: most image models cannot render Chinese characters correctly. For embedded text on a CJK-language deck, either (a) use English equivalents in the image and leave Chinese to SVG overlays, or (b) accept malformed glyphs and regenerate manually.
+**CJK note**: most image models render Chinese characters poorly. For embedded text on a CJK deck, prefer English in the image or accept malformed glyphs.
### 5.4 No brand names or trademarks in the subject
> The image must not depict identifiable brand logos, trademarks, or product likenesses unless the row's Reference explicitly names a real brand asset the user owns.
-### 5.5 Deck-wide visual coherence
-
-Every AI image in one deck shares **one** rendering and **one** palette. Mixing renderings across images in the same deck destroys visual coherence. If a single image truly needs a different rendering (e.g. a corporate-photo team shot alongside otherwise vector-illustration imagery), record this as an exception in the row's `Reference` and isolate it to that one image.
-
---
## 6. Manifest Schema
@@ -385,18 +385,20 @@ If Path A's backend fails twice in a row:
## 8. Common Issues & Variant Workflow
-### Default Inference When No `Reference` Provided
+### Reference field is blank — quick examples
+
+When the Resource List row has no `Reference`, infer a reasonable image from `Purpose`. Examples (not prescriptions):
-| Purpose | Default Inference (assembled from rendering + palette + type) |
-|---------|---------------------------------------------------------------|
-| Cover | `type: background` or `hero`; `page_role: hero_page` when the cover is image-led; `text_policy: embedded` when title is part of the design, else `none` |
-| Chapter divider | `type: background`; `page_role: hero_page` typical; `text_policy: embedded` when chapter title is part of the design |
-| Methodology / framework illustration | `type: framework`, `text_policy: none` — SVG carries node labels |
-| Process / workflow illustration | `type: flowchart`, `text_policy: none` — SVG carries step labels |
-| Before/After or two-option page | `type: comparison`, `text_policy: none` or `embedded` |
-| Team / lifestyle photo | `type: scene`, `text_policy: none`, rendering should be `corporate-photo` or `warm-scene` |
-| Big-number / hero quote block | `type: typography` or `hero`; `page_role: hero_page` typical; `text_policy: embedded` (the number/quote is the visual) |
-| Mood transition / atmosphere | `type: scene` or `background`; `page_role: hero_page`; `text_policy: none` (or decorative lettering if the mood calls for it) |
+| Purpose | A reasonable starting point |
+|---------|-----------------------------|
+| Cover | `type: background` or `hero`; choose page_role and text_policy by what the cover should communicate |
+| Chapter divider | `type: background` or `hero`; often `hero_page` with a designed chapter title |
+| Methodology / framework illustration | `type: framework` |
+| Process / workflow illustration | `type: flowchart` |
+| Before/After or two-option page | `type: comparison` |
+| Team / lifestyle photo | `type: scene`; rendering = `corporate-photo` or `warm-scene` |
+| Big-number / hero quote block | `type: typography` or `hero`; often `hero_page` |
+| Mood transition / atmosphere | `type: scene` or `background`; often `hero_page` |
### When Images Are Unsatisfactory
@@ -409,7 +411,7 @@ Diagnose the failure category, adjust the **one specific dimension** responsible
| Colors don't match deck | HEX not echoed in prompt, or palette proportion rule omitted | Repeat HEX values 2-3 times in the prompt; restate palette proportion rule |
| Hex code or color name visible as text in image | Missing §5.1 closing sentence | Append the §5.1 hard rule verbatim |
| Garbled letters in supposedly text-free image | `text_policy: none` rule too weak | Strengthen with explicit list: "no letters, no numbers, no words, no signs, no labels, no captions, no watermarks" |
-| Composition too busy, no room for SVG overlay | Missing container reminder | Add explicit "leaves at least 30% negative space in the {center / left third / lower band} for text overlay" |
+| SVG text overlay clashes with busy image area | Page design needs negative space the prompt didn't request | Add a composition cue like "leave the {center / left third / lower band} relatively calm for text overlay" — only when the page actually overlays text on top of the image |
| Subject vague | Reference field too abstract | Rewrite reference with concrete nouns (verbs + objects) |
| Faces too realistic / uncanny | §5.2 rule omitted, or rendering is photo-incompatible | Either append §5.2, or switch rendering to a non-photo family |
diff --git a/skills/ppt-master/references/image-type-templates/_index.md b/skills/ppt-master/references/image-type-templates/_index.md
index 55cace04..68bd0f65 100644
--- a/skills/ppt-master/references/image-type-templates/_index.md
+++ b/skills/ppt-master/references/image-type-templates/_index.md
@@ -36,25 +36,25 @@ Each type has its own file with: composition skeleton (LAYOUT / ELEMENTS / NEGAT
For each row in `design_spec.md §VIII Image Resource List`, match `Purpose` against this table.
-| `Purpose` keyword | Type | Suggested `text_policy` |
-|---|---|---|
-| Cover background / chapter background / divider bg | `background` | `none`, or `embedded` (decorative lettering / designed title) when cover/divider is image-led |
-| Product reveal / hero / headline visual | `hero` | `none`, or `embedded` when the headline itself is the visual |
-| Team headshot / speaker bio / founder profile / testimonial | `portrait` | `none` |
-| Big number / KPI single visual / quote | `typography` | `embedded` — the text is the visual |
-| Data summary / metrics rundown / step list | `infographic` | `none` (labels via SVG) or `embedded` (keywords inside the image) |
-| Process / workflow / pipeline / steps with arrows | `flowchart` | `none` (SVG carries step labels) |
-| Methodology / model / framework / architecture diagram | `framework` | `none` (SVG carries node labels) |
-| 2×2 quadrant / SWOT / BCG / Eisenhower / Ansoff | `matrix` | `none`, or `embedded` for axis labels when intentional |
-| Closed-loop process / PDCA / flywheel / continuous improvement | `cycle` | `none` |
-| Conversion funnel / sales pipeline / hiring funnel | `funnel` | `none` |
-| Hierarchy / Maslow / value stack / capability layer | `pyramid` | `none` |
-| Comparison / Before-After / A/B / VS | `comparison` | `none` or `embedded` |
-| History / evolution / roadmap / timeline | `timeline` | `none` (SVG carries milestone labels) |
-| Offices / market presence / regions / supply chain / geography | `map` | `none` |
-| Team / lifestyle / story / scenario / case | `scene` | `none`, or `embedded` for decorative atmosphere lettering |
-
-> Suggested values; per-image judgment wins. Pick `embedded` when in-image text serves the goal (decorative lettering, designed title, hand-lettered keywords, infographic labels). Pick `none` when SVG text is more flexible or the page just needs imagery. Body copy / data points / long quotes never go inside the image regardless of type.
+| `Purpose` keyword | Type |
+|---|---|
+| Cover background / chapter background / divider bg | `background` |
+| Product reveal / hero / headline visual | `hero` |
+| Team headshot / speaker bio / founder profile / testimonial | `portrait` |
+| Big number / KPI single visual / quote | `typography` |
+| Data summary / metrics rundown / step list | `infographic` |
+| Process / workflow / pipeline / steps with arrows | `flowchart` |
+| Methodology / model / framework / architecture diagram | `framework` |
+| 2×2 quadrant / SWOT / BCG / Eisenhower / Ansoff | `matrix` |
+| Closed-loop process / PDCA / flywheel / continuous improvement | `cycle` |
+| Conversion funnel / sales pipeline / hiring funnel | `funnel` |
+| Hierarchy / Maslow / value stack / capability layer | `pyramid` |
+| Comparison / Before-After / A/B / VS | `comparison` |
+| History / evolution / roadmap / timeline | `timeline` |
+| Offices / market presence / regions / supply chain / geography | `map` |
+| Team / lifestyle / story / scenario / case | `scene` |
+
+`text_policy` and `page_role` are decided per image — see each type file's variants section and the page's communication goal.
---
diff --git a/skills/ppt-master/references/image-type-templates/background.md b/skills/ppt-master/references/image-type-templates/background.md
index 84990998..a7b26833 100644
--- a/skills/ppt-master/references/image-type-templates/background.md
+++ b/skills/ppt-master/references/image-type-templates/background.md
@@ -41,7 +41,7 @@ Atmospheric backdrop with **no central subject**. The image's job is to set tone
## 3. Text-policy variants
-### `text_policy: none` (default)
+### `text_policy: none`
Backgrounds are atmospheric — for most uses, keep text out and let SVG handle overlay copy.
diff --git a/skills/ppt-master/references/image-type-templates/cycle.md b/skills/ppt-master/references/image-type-templates/cycle.md
index 5cb67bd9..d137d830 100644
--- a/skills/ppt-master/references/image-type-templates/cycle.md
+++ b/skills/ppt-master/references/image-type-templates/cycle.md
@@ -42,15 +42,15 @@ A **closed-loop process** — 3-6 steps arranged in a circle (or rounded loop),
## 3. Text-policy variants
-### 3.1 `text_policy: none` (default)
+### 3.1 `text_policy: none`
-Step labels are added later as SVG overlay. Each step shows an icon only; the SVG positions the label adjacent to its step. This is the recommended path.
+Each step shows an icon only; step labels are handled in SVG.
Sample fragment:
> NO text, letters, numbers, or step labels in the image. Each step node contains only one simple iconic symbol; SVG text overlay will add step names externally.
-### 3.2 `text_policy: embedded` (when step names are part of the design)
+### 3.2 `text_policy: embedded`
Self-contained cycle diagram with step names typeset into the artwork. Keep step names to single English words ("PLAN", "DO", "CHECK", "ACT") in a font family echoing the deck's body typography.
diff --git a/skills/ppt-master/references/image-type-templates/framework.md b/skills/ppt-master/references/image-type-templates/framework.md
index 6b7c8ec5..14de3131 100644
--- a/skills/ppt-master/references/image-type-templates/framework.md
+++ b/skills/ppt-master/references/image-type-templates/framework.md
@@ -70,20 +70,15 @@ Inner padding: 15-18% on all sides. Framework imagery suffers most when satellit
## 3. Text-policy variants
-### `text_policy: none` (default — recommended for most decks)
+### `text_policy: none`
-The image shows **only the geometric structure**. Labels for each satellite / cell / layer are added in SVG overlay around the image.
-
-This is the recommended path because:
-- SVG labels are editable, translatable, accurate
-- The model can fail on text rendering; geometry it can do
-- The page looks more polished with crisp vector text overlaying the AI art
+The image shows the geometric structure with iconic symbols only; labels are handled in SVG.
Sample fragment to add to the prompt:
-> Pure geometric structure — no labels, no captions, no text, no letters or numbers anywhere in the image. Each satellite/cell/layer contains a simple iconic symbol only (gear, arrow, chart, etc.). SVG text overlay will be added externally.
+> Pure geometric structure — no labels, no captions, no text, no letters or numbers anywhere in the image. Each satellite/cell/layer contains a simple iconic symbol only (gear, arrow, chart, etc.).
-### `text_policy: embedded` (rare — for sketch-notes / ink-notes hand-lettered decks)
+### `text_policy: embedded`
A short keyword (1-2 English words) appears inside or beside each satellite / cell / layer.
diff --git a/skills/ppt-master/references/image-type-templates/funnel.md b/skills/ppt-master/references/image-type-templates/funnel.md
index 8635f6e5..f0145312 100644
--- a/skills/ppt-master/references/image-type-templates/funnel.md
+++ b/skills/ppt-master/references/image-type-templates/funnel.md
@@ -43,15 +43,15 @@ A **top-wide, bottom-narrow stack** of horizontal bands representing successive
## 3. Text-policy variants
-### 3.1 `text_policy: none` (default)
+### 3.1 `text_policy: none`
-Stage labels are added later as SVG overlay. Each band shows an icon only; the SVG places stage names beside or inside each band. Recommended path.
+Each band shows an icon only; stage labels are handled in SVG.
Sample fragment:
> NO text, letters, numbers, or stage labels in the image. Each band contains only one simple iconic symbol; SVG text overlay will add stage names externally.
-### 3.2 `text_policy: embedded` (when stage names are part of the design)
+### 3.2 `text_policy: embedded`
Self-contained funnel diagram where band names are typeset into the artwork. Keep band names to single English words ("AWARE", "LIKE", "BUY", "REFER") in a font family echoing the deck's body typography. High failure risk on 5+ band funnels — stay at 3-4 bands when going embedded.
diff --git a/skills/ppt-master/references/image-type-templates/hero.md b/skills/ppt-master/references/image-type-templates/hero.md
index e8850af6..09402d74 100644
--- a/skills/ppt-master/references/image-type-templates/hero.md
+++ b/skills/ppt-master/references/image-type-templates/hero.md
@@ -44,7 +44,7 @@ Single dominant subject occupying 60-70% of the canvas. The hero image *anchors*
The hero subject is the visual; any title or label comes from SVG overlay.
-### `text_policy: embedded` (when the hero needs in-image text)
+### `text_policy: embedded`
The hero subject itself includes text — product name on packaging, a hand-lettered word as part of the subject, a designed title floating beside the figure. Specify font family in the prompt to echo the deck's body typography. If the headline *is* the entire visual, switch to `typography` type.
diff --git a/skills/ppt-master/references/image-type-templates/map.md b/skills/ppt-master/references/image-type-templates/map.md
index 0f91b999..de4fcc20 100644
--- a/skills/ppt-master/references/image-type-templates/map.md
+++ b/skills/ppt-master/references/image-type-templates/map.md
@@ -43,7 +43,7 @@ A **stylized geographic outline** (country / region / continent / world) with an
## 3. Text-policy variants
-### 3.1 `text_policy: none` (default)
+### 3.1 `text_policy: none`
City / country / region names are added later as SVG overlay. The map shows markers only; SVG text labels each marker position.
@@ -51,7 +51,7 @@ Sample fragment:
> NO text, letters, numbers, country names, city names, or marker labels in the image. Markers are dots/pins only; SVG text overlay will add all labels externally.
-### 3.2 `text_policy: embedded` (when place names belong in the artwork)
+### 3.2 `text_policy: embedded`
Self-contained reference map where place names are typeset into the design. High failure risk — image models often misspell place names or place them at wrong coordinates. Prefer `none` + SVG overlay unless the design language genuinely requires in-image labels (vintage cartography, atlas-style poster).
diff --git a/skills/ppt-master/references/image-type-templates/matrix.md b/skills/ppt-master/references/image-type-templates/matrix.md
index 342837ea..224864d5 100644
--- a/skills/ppt-master/references/image-type-templates/matrix.md
+++ b/skills/ppt-master/references/image-type-templates/matrix.md
@@ -44,15 +44,15 @@ A **2×2 quadrant grid** (occasionally 3×3) where each cell carries its own lab
## 3. Text-policy variants
-### 3.1 `text_policy: none` (default)
+### 3.1 `text_policy: none`
-Quadrants carry icons only; axis labels and quadrant names are added later as SVG overlay. This is the safer choice — image models often render axis-label text inconsistently across the four positions.
+Quadrants carry icons only; axis labels and quadrant names are handled in SVG. Image models often render axis-label text inconsistently across the four positions, so this is the lower-risk path when label precision matters.
Sample fragment:
> NO text, letters, numbers, axis labels, or quadrant names in the image. Each quadrant contains a simple iconic symbol only; SVG text overlay will add axis labels and quadrant names externally.
-### 3.2 `text_policy: embedded` (when axis labels are part of the design)
+### 3.2 `text_policy: embedded`
When the matrix's identity comes from its in-image lettering — a stylized SWOT poster with the four letters as visuals, a designer-matrix where axis labels are typeset into the artwork. Keep labels to single English words ("HIGH", "LOW", "GROW", "HOLD"). Specify the font family in the prompt to echo the deck's body typography.
diff --git a/skills/ppt-master/references/image-type-templates/portrait.md b/skills/ppt-master/references/image-type-templates/portrait.md
index c8310ad5..0c029d3f 100644
--- a/skills/ppt-master/references/image-type-templates/portrait.md
+++ b/skills/ppt-master/references/image-type-templates/portrait.md
@@ -44,15 +44,17 @@ A **single person headshot or upper-body shot** — frontal or three-quarter tur
## 3. Text-policy variants
-### 3.1 `text_policy: none` (default)
+### 3.1 `text_policy: none`
-Names, titles, and quotes are added later as SVG overlay. The portrait shows the person only.
+The portrait shows the person only; names, titles, and quotes are handled in SVG.
Sample fragment:
-> NO text, letters, numbers, name tags, or captions in the image. The portrait is the person only; SVG text overlay will add name, title, and any quote externally.
+> NO text, letters, numbers, name tags, or captions in the image. The portrait is the person only.
-### 3.2 `text_policy: embedded` is essentially never used for this type — switch to `hero` or `typography` if text is required as part of the image.
+### 3.2 `text_policy: embedded`
+
+If text is part of the page's visual (a name carved as a design element, a quote alongside the figure), the type is likely `hero` or `typography` rather than `portrait`.
---
diff --git a/skills/ppt-master/references/image-type-templates/pyramid.md b/skills/ppt-master/references/image-type-templates/pyramid.md
index 1bae460a..c796e87b 100644
--- a/skills/ppt-master/references/image-type-templates/pyramid.md
+++ b/skills/ppt-master/references/image-type-templates/pyramid.md
@@ -42,7 +42,7 @@ A **tiered triangular stack** — bottom-wide, top-narrow — where each layer c
## 3. Text-policy variants
-### 3.1 `text_policy: none` (default)
+### 3.1 `text_policy: none`
Tier labels are added later as SVG overlay. Each tier shows an icon only; the SVG places tier names beside or inside.
@@ -50,7 +50,7 @@ Sample fragment:
> NO text, letters, numbers, or tier labels in the image. Each tier contains only one simple iconic symbol; SVG text overlay will add tier names externally.
-### 3.2 `text_policy: embedded` (when tier names are part of the design)
+### 3.2 `text_policy: embedded`
Self-contained pyramid with tier names typeset into the artwork. Keep tier names to single English words in a font family echoing the deck's body typography. High failure risk on 5+ tier pyramids — stay at 3-4 tiers when going embedded.
diff --git a/skills/ppt-master/references/image-type-templates/scene.md b/skills/ppt-master/references/image-type-templates/scene.md
index 94e68418..5d93c95a 100644
--- a/skills/ppt-master/references/image-type-templates/scene.md
+++ b/skills/ppt-master/references/image-type-templates/scene.md
@@ -40,7 +40,7 @@ Atmospheric environment with narrative — a moment, a place, a situation render
## 3. Text-policy variants
-### `text_policy: none` (default)
+### `text_policy: none`
Scenes are visual narratives — most read better with text routed to SVG overlay.
diff --git a/skills/ppt-master/references/image-type-templates/typography.md b/skills/ppt-master/references/image-type-templates/typography.md
index 7ad2d74b..4496ca22 100644
--- a/skills/ppt-master/references/image-type-templates/typography.md
+++ b/skills/ppt-master/references/image-type-templates/typography.md
@@ -63,9 +63,9 @@ When the desired headline is long or CJK: switch to `background` type and overla
## 4. Text-policy variants
-### `text_policy: embedded` (required)
+### `text_policy: embedded`
-typography type is the **one type where embedded text is the entire point**. Always use `text_policy: embedded`.
+typography type is built around in-image text — the headline word, number, or quote is the visual. `text_policy: embedded` is the natural choice.
Sample fragment:
@@ -73,9 +73,9 @@ Sample fragment:
**Font family in prompt**: name the family so the lettering coheres with the deck's SVG typography. Read `spec_lock.md deck_typography` if present; otherwise infer from rendering (vector-illustration / flat → clean sans-serif; editorial → serif; sketch-notes / ink-notes → hand-written; 3d-isometric / digital-dashboard → geometric display).
-### `text_policy: none` (essentially never)
+### `text_policy: none`
-If text_policy is none, this isn't typography — switch to `background` or `hero`.
+If the page wants no text inside the image, the type is probably `background` or `hero`, not `typography`.
---
diff --git a/skills/ppt-master/references/strategist.md b/skills/ppt-master/references/strategist.md
index 3e7ee986..aa059f20 100644
--- a/skills/ppt-master/references/strategist.md
+++ b/skills/ppt-master/references/strategist.md
@@ -383,31 +383,7 @@ Image_Generator reads these two fields and applies them deck-wide. If both are a
#### hero_page suggestion (same confirmation turn)
-After the user picks a candidate, scan the outline for pages where the image speaks louder than text — covers, chapter dividers, mood transitions, single-number data heroes, closing quotes, brand summaries. Surface a suggestion list:
-
-```
-Suggested hero pages (image as page's main voice; SVG overlay minimal):
- Slide 1 — Cover (image-led)
- Slide 8 — Chapter divider: "Market Forces"
- Slide 17 — Data hero: the 73% number
- Slide 24 — Closing
-Confirm, edit, or reply "all local" to skip.
-```
-
-User confirmation rules:
-- Accept all / pick subset / add others / reject all
-- Result is recorded as `page_role: hero_page` on the matching `ai` rows in the resource list
-- Density is judgment-based — image-heavy brand decks may use more, data/teaching decks fewer. No fixed quota.
-
-#### Typography hint for in-image text
-
-When any hero_page or other `text_policy: embedded` row exists, also record a font family hint in `spec_lock.md` so Image_Generator's prompts echo the deck's SVG typography:
-
-```
-- deck_typography: clean sans-serif # or: serif editorial / hand-written / geometric display
-```
-
-Family-name granularity is enough — actual font choice happens in Executor's SVG; the hint exists so the AI image's lettering doesn't clash with the deck.
+After the user picks a candidate, scan the outline and surface any pages where the image makes more sense as the page's main voice than as a local block. Present them as a short list and let the user confirm, edit, or skip. Result is recorded as `page_role: hero_page` on the matching `ai` rows. Density is judgment-based — no fixed quota.
**When selection includes B**, you must run `python3 scripts/analyze_images.py <project_path>/images` before outputting the spec, and integrate scan results into the image resource list.
@@ -424,8 +400,8 @@ Family-name granularity is enough — actual font choice happens in Executor's S
| **Acquire Via** | `ai` / `web` / `user` / `placeholder` — drives Step 5 dispatch |
| Status | Initial status must be `Pending`, `Existing`, or `Placeholder`; see [`svg-image-embedding.md`](svg-image-embedding.md) for the full status enum |
| **Reference** | Free-form **intent description** (NOT a search query); feeds Image_Generator (ai) or Image_Searcher (web) |
-| `text_policy` (optional, `ai` rows only) | `none` (default — no text in image, SVG overlays labels) or `embedded` (image contains decorative lettering, a designed title, or hand-lettered keywords as part of the artwork). Leave blank for default. Long body / data / lists never go inside the image. |
-| `page_role` (optional, `ai` rows only) | `local` (default — image is a region block on an SVG page) or `hero_page` (image is the page's main voice; SVG overlay minimal or empty). Set per the hero_page suggestion confirmed in h.5. Same rendering and palette as the rest of the deck regardless. |
+| `text_policy` (optional, `ai` rows only) | `none` (no text in image) or `embedded` (text is part of the artwork). Leave blank when Image_Generator should decide per row. Long body / data / lists stay in SVG. |
+| `page_role` (optional, `ai` rows only) | `local` (image is a region block on an SVG page) or `hero_page` (image is the page's main voice). Leave blank when Image_Generator should decide per row. |
**No-crop flag (exception only)**: most images are croppable — Executor defaults to `preserveAspectRatio="xMidYMid slice"`. When an image must NOT lose pixels (data screenshots, charts, certificates, contracts, dense diagrams), append `no-crop` to its `spec_lock.md images` entry. Executor will then size the container to the native ratio and use `meet`. Don't tag the rest.