Skill: Google Slides Export

Purpose

Create properly formatted Google Slides presentations via the MCP API. Prevents API errors and ensures every deck looks professional -- no text overflow, no overlapping elements, consistent brand design.

When to Apply

Automatically whenever:

• mcp__google-workspace__batch_update_presentation will be called
• A Google Slides deck is being designed or built
• The google-slides-creator or google-slides-reviewer agent is running

---

Section A: Pre-Flight Checklist

Run this checklist BEFORE building any batch_update requests. Fix all violations first.

• Object IDs >= 5 characters -- every NEW object ID you create must be 5+ chars OK: slide_2, hdr_bg, title_box, body_txt BAD: s2, t1, bg, p_t1 (Only applies to IDs you define; existing IDs from the API can be any length)
• No zero-weight outlines -- never set outline.weight.magnitude = 0 BAD: "weight": {"magnitude": 0, "unit": "PT"} -> API error OK: Omit the updateShapeProperties outline block entirely (no border = default)
• Slide creation uses createSlide -- not addSlide (not a valid request type)
• Batches <= 50 requests -- split into batches of 50 and apply sequentially Each batch must complete before the next starts (do not parallelize)
• deleteObject only for confirmed IDs -- only delete element IDs returned by get_page or get_presentation; never delete IDs you invented
• Text boxes are generous -- every text box that holds variable content must be at least 3000000 EMU wide and use autoFit: AUTO_FIT

---

Section B: Design System

Page Dimensions

Width:  9144000 EMU  (10 inches)
Height: 5143500 EMU  (5.625 inches)
1 inch = 914400 EMU

Color Palette (Default Brand)

Dark navy:    {red: 0.118, green: 0.161, blue: 0.231}   -- header bars, full-bg dividers
Orange:       {red: 0.851, green: 0.467, blue: 0.024}   -- highlights, accent callouts
Accent green: {red: 0.02,  green: 0.588, blue: 0.412}   -- secondary metrics
Off-white:    {red: 0.969, green: 0.965, blue: 0.949}   -- slide background
White:        {red: 1.0,   green: 1.0,   blue: 1.0}     -- card backgrounds, text on dark
Dark text:    {red: 0.133, green: 0.133, blue: 0.133}   -- body text on light backgrounds

Layout Zones

Header bar:       x=0,       y=0,       w=9144000, h=686000
Title in header:  x=457200,  y=100000,  w=8229600, h=486000
Content start:    y=800000
Left margin:      x=457200   (0.5 inch)
Right edge:       x=8686800
Content width:    8229600    (9 inches usable)
Bottom safe zone: y <= 4800000 (leaves 0.37in gutter above slide bottom)

Formatting Rules

1. autoFit on all content text boxes -- use "contentAlignment": "TOP" + "autoFit": {"autoFitType": "AUTO_FIT"} in updateShapeProperties. Never set a fixed height on a text box that holds variable text content.

2. Minimum text box width: 3000000 EMU (3.3 inches) for any box holding sentences. Narrow boxes (metric labels, short tags) may be smaller.

3. Font size hierarchy:

Role	Size	Weight
Full-slide section title (dark bg)	28-32pt	Bold
Slide title (in header bar)	20-24pt	Bold
Insight headline (below header)	14-16pt	Bold
Body / bullet text	11-13pt	Regular
Metric value (KPI card)	24-30pt	Bold
Card label / footnote	9-11pt	Regular

4. Maximum 5 content elements per slide (not counting the header RECTANGLE and title TEXT_BOX). If you need more, split into two slides.

5. Maximum 3 bullet points per text box. Max 70 characters per bullet. If content is longer, split into two slides or use a two-column layout.

6. Card pattern for KPI metrics:

• Create a white RECTANGLE as the card background
• Layer a large metric value TEXT_BOX on top
• Layer a smaller label TEXT_BOX below the value
• Card objects must be created in order: background RECTANGLE first, then text boxes

7. Paragraph spacing inside text boxes:

"paragraphStyle": {
  "spaceAbove": {"magnitude": 4, "unit": "PT"},
  "spaceBelow": {"magnitude": 4, "unit": "PT"},
  "lineSpacing": 115
}

8. Never use spaces or tabs for visual alignment. Use separate text boxes positioned at the correct x coordinate.

---

Section C: Layout Library

Five pre-validated slide type recipes. Each specifies exact createShape / createSlide requests. Always pick the closest matching type; fill in text content only.

Object ID naming convention: {type}_{slidenum}_{role} Examples: slide_2, hdr_2, ttl_2, hdl_2, bdy_2, cd1_2, cv1_2, cl1_2

---

Type 1 -- Title Slide

Use for: deck opening slide Elements: 3

1. RECTANGLE (full-background, dark navy)
  objectId: "bg_{n}"
  size:      w=9144000, h=5143500
  position:  x=0, y=0
  fill:      dark navy

2. TEXT_BOX (main title)
  objectId: "ttl_{n}"
  size:      w=8229600, h=900000
  position:  x=457200, y=1700000
  text:      [DECK TITLE -- headline, not label]
  font:      28pt, white, bold
  autoFit:   AUTO_FIT

3. TEXT_BOX (subtitle / context)
  objectId: "sub_{n}"
  size:      w=8229600, h=500000
  position:  x=457200, y=2700000
  text:      [Scope | Date range | Author]
  font:      16pt, off-white, regular
  autoFit:   AUTO_FIT

---

Type 2 -- Section Divider

Use for: transitions between major sections Elements: 2

1. RECTANGLE (full-background, dark navy)
  objectId: "bg_{n}"
  size:      w=9144000, h=5143500
  position:  x=0, y=0
  fill:      dark navy

2. TEXT_BOX (section name)
  objectId: "sec_{n}"
  size:      w=8229600, h=700000
  position:  x=457200, y=1900000
  text:      [SECTION NAME]
  font:      28pt, white, bold, centered
  autoFit:   AUTO_FIT

---

Type 3 -- Header + Bullets

Use for: insight findings, recommendation lists, methodology, any text-heavy slide Elements: 4

1. RECTANGLE (header bar, dark navy)
  objectId: "hdr_{n}"
  size:      w=9144000, h=686000
  position:  x=0, y=0
  fill:      dark navy

2. TEXT_BOX (slide title, inside header)
  objectId: "ttl_{n}"
  size:      w=8229600, h=486000
  position:  x=457200, y=100000
  text:      [SLIDE TITLE -- max 60 chars]
  font:      20pt, white, bold

3. TEXT_BOX (insight headline)
  objectId: "hdl_{n}"
  size:      w=8229600, h=450000
  position:  x=457200, y=800000
  text:      [INSIGHT HEADLINE -- the "so what", max 100 chars]
  font:      14pt, dark navy, bold
  autoFit:   AUTO_FIT

4. TEXT_BOX (body bullets)
  objectId: "bdy_{n}"
  size:      w=8229600, h=3300000
  position:  x=457200, y=1350000
  text:      [bullet 1\nbullet 2\nbullet 3]  (max 3 bullets, 70 chars each)
  font:      12pt, dark text, regular
  autoFit:   AUTO_FIT

---

Type 4 -- KPI Cards

Use for: metric dashboards with 2-4 key numbers side by side Max 4 cards. Elements: header (2) + per card (3 each) = 5-14 total.

Card x-positions by count:

2 cards: card_w=3800000  x1=457200,  x2=4686800
3 cards: card_w=2500000  x1=457200,  x2=3214200,  x3=5971200
4 cards: card_w=1800000  x1=457200,  x2=2514200,  x3=4571200,  x4=6628200

Card height: 2000000 EMU. Card y-position: 900000.

1. RECTANGLE (header bar, dark navy)    -- objectId: "hdr_{n}"
2. TEXT_BOX (slide title in header)     -- objectId: "ttl_{n}"
  [same as Type 3 header elements]

Per card i (repeat for each metric):
3i. RECTANGLE (card background, white)
   objectId: "cd{i}_{n}"
   size:      w={card_w}, h=2000000
   position:  x={xi}, y=900000
   fill:      white, rounded corners (optional)

4i. TEXT_BOX (metric value)
   objectId: "cv{i}_{n}"
   size:      w={card_w - 200000}, h=900000
   position:  x={xi + 100000}, y=1050000
   text:      [METRIC VALUE -- e.g., "40.2M" or "39.4%"]
   font:      28pt, dark navy, bold, centered

5i. TEXT_BOX (metric label)
   objectId: "cl{i}_{n}"
   size:      w={card_w - 200000}, h=400000
   position:  x={xi + 100000}, y=2050000
   text:      [METRIC LABEL -- e.g., "Monthly Active Users"]
   font:      10pt, dark text, regular, centered
   autoFit:   AUTO_FIT

---

Type 5 -- Two-Column

Use for: side-by-side comparisons, before/after, two segments Elements: 4

1. RECTANGLE (header bar, dark navy)    -- objectId: "hdr_{n}"
2. TEXT_BOX (slide title in header)     -- objectId: "ttl_{n}"
  [same as Type 3 header elements]

3. TEXT_BOX (left column)
  objectId: "lcl_{n}"
  size:      w=3900000, h=3700000
  position:  x=457200, y=900000
  font:      12pt, dark text, regular
  autoFit:   AUTO_FIT

4. TEXT_BOX (right column)
  objectId: "rcl_{n}"
  size:      w=3900000, h=3700000
  position:  x=4786800, y=900000
  font:      12pt, dark text, regular
  autoFit:   AUTO_FIT

---

Type 6 -- Chart Slide (Header + Text + Image)

Use for: findings with an embedded chart image Elements: 5

1. RECTANGLE (header bar, dark navy)    -- objectId: "hdr_{n}"
2. TEXT_BOX (slide title in header)     -- objectId: "ttl_{n}"
  [same as Type 3 header elements]

3. TEXT_BOX (insight headline)
  objectId: "hdl_{n}"
  size:      w=4000000, h=450000
  position:  x=457200, y=800000
  text:      [INSIGHT HEADLINE -- the "so what", max 100 chars]
  font:      14pt, dark navy, bold
  autoFit:   AUTO_FIT

4. TEXT_BOX (body text, left side)
  objectId: "bdy_{n}"
  size:      w=4000000, h=3300000
  position:  x=457200, y=1350000
  text:      [3 bullets max, 70 chars each]
  font:      12pt, dark text, regular
  autoFit:   AUTO_FIT

5. IMAGE (chart, right side)
  objectId: "img_{n}"
  size:      w=4300000, h=3500000
  position:  x=4572000, y=1300000
  url:       [HTTPS URL -- see Image Insertion workflow below]

Image Insertion Workflow (for Type 6 slides):

Charts are local PNG files that must be reachable at an HTTPS URL before insertion. The Google Slides API does NOT accept data:image/... data URIs.

Data handling rule: charts go only to the user's own Google Drive. Never upload charts or data to public file hosts; the analysis may contain confidential numbers. Drive files stay under the user's control, with just enough link access for the Slides API to fetch the image.

1. Upload each chart PNG to Drive using the chart-to-drive skill (.claude/skills/chart-to-drive/SKILL.md). It uploads via the Google MCP tools, sets link access, and returns a direct-download URL of the form https://drive.google.com/uc?id=<FILE_ID> that works in createImage.

2. Insert into slides via createImage:

{"createImage": {
  "objectId": "img_{n}",
  "url": "<dl_url from step 1>",
  "elementProperties": {
    "pageObjectId": "slide_{n}",
    "size": {"width": {"magnitude": 4300000, "unit": "EMU"},
             "height": {"magnitude": 3500000, "unit": "EMU"}},
    "transform": {"scaleX": 1, "scaleY": 1,
                   "translateX": 4572000, "translateY": 1300000, "unit": "EMU"}
  }
}}

3. Delete the placeholder frame (if one exists):

{"deleteObject": {"objectId": "cpf_{n}"}}

4. The chart is already in Drive (step 1), so no extra save step is needed; record the Drive file ID in the chart map for reuse.

Important API constraints:

• BAD: data:image/png;base64,... -> returns 403 Forbidden
• BAD: http://localhost:... -> MCP server cannot access local URLs
• OK: a Drive direct-download URL (https://drive.google.com/uc?id=...) serving the image
• All 5 createImage + deleteObject requests can go in a single batch

---

Section D-1: Data Stamp & Speaker Notes Provenance

Data Stamp Text Box

For finding slides (Type 3 and Type 6), add a data stamp text box at the bottom-right corner. This provides at-a-glance provenance for every claim.

TEXT_BOX (data stamp)
  objectId: "dst_{n}"
  size:      w=4000000, h=250000
  position:  x=4686800, y=4800000
  text:      [abbreviated stamp: "50K | Jan-Mar 2026 | EVENTS | B (82)"]
  font:      8pt, {red: 0.6, green: 0.6, blue: 0.6}, regular
  alignment: RIGHT
  autoFit:   AUTO_FIT

When to include: Any slide presenting a specific data finding (insight headline + evidence). When to skip: Title slides, section dividers, KPI dashboards, recommendation lists, appendix.

The abbreviated data stamp format is: {row_count} | {date_range} | {primary_table} | {grade} ({score})

Built via helpers/provenance/provenance_assembler.py:

from helpers.provenance.provenance_assembler import render_data_stamp
stamp_text = render_data_stamp(block["data_stamp"], level="abbreviated")

Speaker Notes Provenance Format

For finding slides, speaker notes should contain full provenance so the presenter can answer "where does this number come from?" questions:

Data: [145K rows | Jan-Mar 2026 | ORDERS | Confidence: B (82/100)]
Methodology: segmented comparison, SUM by segment
SQL: SELECT segment, SUM(revenue) FROM orders GROUP BY segment
Verification: Type B: Parts-to-whole — Verified (PASS, diff 0.2%)

Insert via insertText on the slide's notes page object.

---

Section D: Reference Decks

To verify design decisions, use decks previously created by this pipeline as visual references. Add your own reference deck IDs here after creating decks you're happy with.

Visual patterns to follow:

• Card-based layouts: colored RECTANGLE backgrounds with text boxes layered on top
• Section dividers: full dark-navy background with large centered text only
• Minimal element count: 2-4 elements on dividers, 5-10 on content slides
• Flow diagrams: CURVED_LEFT_ARROW shapes for process flows (not table data)
• Round rectangles (ROUND_RECTANGLE) as alternative card containers