/oci

OCI PPTX Skill

Do NOT load this skill when

Do not load this skill for unrelated general programming, non-Oracle cloud work, or questions covered by a narrower sibling skill.

When to Use

Load this skill for: the user asks to "create Oracle slides", "edit an Oracle deck", "build a CloudWorld presentation", "review Oracle-branded PPTX", or "apply Oracle brand to slides".

Prefer this skill only for its named domain. For broader OCI architecture triage, start with oci/best-practices as the router.

This skill holds Oracle brand and slide-design rules only. It does not bundle PPTX tooling: use python-pptx or PptxGenJS (see references/implementation-examples.md), or your agent's own PPTX skill, for file manipulation.

Oracle-Branded Presentations with Cognitive Science

Quick Reference

Do NOT load multiple references simultaneously.

The Five Cognitive Science Rules

Based on Sweller (cognitive load theory), Mayer (multimedia learning), Phillips (Death by PowerPoint):

1. One Message Per Slide

Each slide = exactly ONE key point. Secondary messages compete for attention and are poorly retained.

2. Avoid Redundancy Effect

Reading dense on-screen text while hearing the same words spoken hurts comprehension (Mayer's redundancy principle). Max 50 words per slide. Move details to speaker notes.

3. Size = Importance Hierarchy

Largest element = most important content (may NOT be the headline). Make key metrics larger than titles when the metric IS the message.

4. Context-Aware Backgrounds

Dark slides make the SPEAKER the highest-contrast object in the room — that's the cognitive science intent.

5. Maximum 6 Objects Per Slide

Count ALL elements: images + shapes + text boxes + charts + decorative (exclude footer/header). Beyond about 6 elements the audience must scan instead of grasp the slide at a glance. Exception: Grid layouts where items form a single visual unit.

Oracle Brand Quick Reference

Typography

Font: Oracle Sans Tab (Regular/Bold)

• Verify installed: fc-list | grep -i "oracle"
• Fallback: Arial (always prefer Oracle Sans Tab)

Quick Colors by Pillar

For full palette details, read references/color-palettes.md.

Backgrounds: Dark #0B2540 / #1a1a1a | Light #FFFFFF / #F5F5F5

Oracle Footer (Mandatory on All Slides)

• Left: Copyright © <current year>, Oracle and/or its affiliates
• Right (optional): omit for public events | | Confidential: Internal | | Confidential: Restricted | | Confidential: Highly Restricted

Design Workflow

Step 1: Context Assessment

1. Where presented? Large live event → Dark primary. Virtual → Mix. Document → Light.
2. Who is audience? Executives: large metrics, minimal text. Technical: more detail, still ≤6 objects.
3. Which pillar? Determines color palette.

Step 2: Per-Slide Content Strategy

1. Identify THE one message
2. Find most important element → make it largest (metric? insight? visual?)
3. Count objects — ≤6?
4. Text audit — >50 words? → move to speaker notes
5. Redundancy check — slide text duplicates speaker notes? → choose one

Step 3: Visual Design

Layout options: Two-column | Half-bleed image | Large metric callout (60-72pt + small label) | Icon + text rows | 2×2 or 2×3 grid

Every slide needs a visual element: Chart, icon set, product screenshot, customer photo, or organic shape motif. Oracle icon style: icons in colored circles.

Oracle Anti-Patterns (NEVER)

• Generic blue defaults — use pillar-specific colors, not default blue
• Centered body text — left-align paragraphs and lists
• Repeated layouts — vary slide structures throughout deck
• Accent lines under titles — hallmark of AI-generated slides, violates Oracle brand
• Text-only slides — every slide needs a visual element
• Multiple messages per slide — confuses audience, kills retention
• More than 6 objects — cognitive overload
• Missing footer — Oracle requires footer on all slides
• Wrong background for context — dark for live events, not documents
• Arial when Oracle Sans Tab available — always check font availability first

Template Rules

Accessibility (Oracle Standard)

Reading order (screen readers read bottom to top in Selection Pane):

1. Home → Arrange → Selection Pane
2. Drag most important content to bottom of list

Alt text: Right-click image → "View Alt Text". Describe content for blind/low-vision users. Format: "Bar chart showing 127% revenue growth YoY" Mark purely decorative elements as decorative.

Contrast: Minimum 4.5:1 (WCAG AA). Oracle palettes pre-selected for strong contrast.

Cognitive Load Validation Checklist

Before finalizing each slide:

• One message — summarizable in one sentence?
• ≤6 objects (images + shapes + text boxes + charts, excluding footer)?
• ≤50 words on slide (verbose content in speaker notes)?
• Largest element = most important content?
• No redundancy between slide text and speaker notes?
• Contrast ≥4.5:1?
• Background matches context (dark live / light docs / mixed virtual)?
• Visual element present?
• Oracle Sans Tab font used?
• Correct pillar color palette?
• Footer present with current year and confidentiality level?
• Alt text on all images?
• Selection Pane reading order correct?

QA Workflow (Required)

Mindset: Assume problems exist. First render is almost never correct.

Phase 1: Content QA

python -m markitdown output.pptx
python -m markitdown output.pptx | grep -iE "xxxx|lorem|ipsum"

Phase 2: Cognitive Load QA

Per slide: object count ≤6? word count ≤50? one clear message? largest = most important?

Phase 3: Brand Compliance QA

Oracle Sans Tab? Correct pillar palette? Background matches context? Footer correct year? No AI patterns (accent lines, centered body text)?

Phase 4: Visual QA with Subagents

# Convert to images for visual inspection
soffice --headless --convert-to pdf output.pptx   # LibreOffice
pdftoppm -jpeg -r 150 output.pdf slide

Subagent prompt:

Visually inspect these Oracle-branded slides. Assume there are issues — find them.
Check: overlapping elements, text overflow, object count >6, low contrast,
text-only slides, wrong font, missing/incorrect footer, wrong background for context.
Report ALL issues per slide.

Verification Loop

1. Generate → convert to images → inspect
2. List ALL issues (cognitive + brand + visual)
3. Fix issues
4. Re-verify affected slides
5. Repeat until no issues found

Do not declare success until at least one fix-and-verify cycle is complete.

Reading/Inspecting Content

python -m markitdown presentation.pptx         # Text extraction
soffice --headless --convert-to pdf presentation.pptx && pdftoppm -jpeg -r 80 presentation.pdf thumb  # Visual check
unzip -o presentation.pptx -d presentation_xml/  # Raw XML inspection (PPTX is a zip)

Arguments

$ARGUMENTS: Optional user-provided target, path, environment, symptom, or constraint. When empty, infer the narrowest safe scope from the current repository context and ask only if multiple high-impact choices remain.