Google Docs Formatter

Google Docs Formatter

Google Docs does NOT render markdown. Text like # Header or **bold** appears as literal characters. Use native Google Docs formatting via the MCP gdrive tools.

Document Location

IMPORTANT: Before using this skill to format documents, ensure the document was created using the /gdrive-manager skill, which enforces the claude-code/ folder structure. All research documents should be in:

• claude-code/experiments/ - Experiment documentation
• claude-code/logs/ - Daily logs
• claude-code/updates/ - Weekly updates
• claude-code/slides/ - Presentations

See .claude/skills/gdrive-manager/references/folder-ids.md for folder IDs.

Key Principle

Always get document content first to find text indices, then apply formatting to those ranges.

Workflow

Step 1: Get Document Content with Indices

mcp__gdrive__getGoogleDocContent(documentId: "your-doc-id")

Returns content with index ranges like:

[1-55] Code vs NL Representation: Mutual Information Analysis
[57-62] TL;DR
[63-189] Code representations consistently encode MORE...

The indices are 1-based and required for all formatting operations.

Step 2: Apply Paragraph Styles (Headings)

mcp__gdrive__formatGoogleDocParagraph(
  documentId: "...",
  startIndex: 1,
  endIndex: 55,
  namedStyleType: "TITLE"  // or HEADING_1, HEADING_2, etc.
)

Available styles:

Alignment options: START, CENTER, END, JUSTIFIED

Step 3: Apply Text Formatting (Bold, Italic, etc.)

mcp__gdrive__formatGoogleDocText(
  documentId: "...",
  startIndex: 191,
  endIndex: 204,
  bold: true
)

Available options:

• bold: true/false
• italic: true/false
• underline: true/false
• strikethrough: true/false
• fontSize: number (in points)
• foregroundColor: { red: 0-1, green: 0-1, blue: 0-1 }

Document Structure

Required: TL;DR Section

Every document MUST start with a TL;DR section immediately after the title. This provides readers with a quick summary before diving into details.

Format:

[Document Title]

TL;DR
• Key finding/point 1
• Key finding/point 2
• Key finding/point 3

[Rest of document...]

Formatting for TL;DR:

1. Apply HEADING_2 style to "TL;DR"
2. Keep it to 2-4 bullet points
3. Each bullet should be one concise sentence
4. Bold the most critical terms/numbers within the TL;DR

Example TL;DR content:

TL;DR
• Code representations encode 1.2 bits MORE mutual information than NL on average
• Effect is consistent across all 6 tested models (p < 0.01)
• Larger models show bigger code vs NL gap

Content Best Practices

Instead of Markdown, Use:

Bullet Points

Use the bullet character directly in content:

• First item
• Second item
• Third item

Tables

Google Docs API has limited table support. Options:

1. Tab-separated text - Simple but no borders
2. Google Sheets - Better for real tables, then link/embed

Tab-separated example:

Model	Code MI	NL MI	Diff
claude-haiku	2.10	0.93	1.17
gpt-4o-mini	1.23	0.35	0.87

Common Formatting Patterns

Format a Research Document

1. Get content indices
2. Apply TITLE to document title
3. Apply HEADING_2 to "TL;DR" section (required, immediately after title)
4. Bold key findings/numbers in the TL;DR bullets
5. Apply HEADING_1 to section headers (Introduction, Methods, Results, etc.)
6. Apply HEADING_2 to subsections
7. Bold key terms and important findings
8. Use bullet points (•) for lists

Bold Key Labels

For configuration sections like:

• Model: Claude Haiku 4.5
• Backend: OpenRouter

Bold "Model:" and "Backend:" by finding their indices and applying bold: true.

Document Search

Find documents by name:

mcp__gdrive__search(query: "Document Name")

Then get the document ID from results to use with formatting tools.

Example: Full Formatting Flow

# 1. Search for document
search_result = mcp__gdrive__search(query="My Research Doc")
doc_id = "extracted-from-search"

# 2. Get content with indices
content = mcp__gdrive__getGoogleDocContent(documentId=doc_id)
# Returns: [1-30] My Research Title\n[32-37] TL;DR\n[39-120] • Key finding...\n[122-135] Introduction...

# 3. Apply title style
mcp__gdrive__formatGoogleDocParagraph(
  documentId=doc_id,
  startIndex=1,
  endIndex=30,
  namedStyleType="TITLE"
)

# 4. Apply HEADING_2 to TL;DR (REQUIRED - always include this)
mcp__gdrive__formatGoogleDocParagraph(
  documentId=doc_id,
  startIndex=32,
  endIndex=37,
  namedStyleType="HEADING_2"
)

# 5. Bold key numbers/findings in TL;DR
mcp__gdrive__formatGoogleDocText(
  documentId=doc_id,
  startIndex=50,  # "1.2 bits" within the TL;DR
  endIndex=58,
  bold=True
)

# 6. Apply heading to main sections
mcp__gdrive__formatGoogleDocParagraph(
  documentId=doc_id,
  startIndex=122,
  endIndex=135,
  namedStyleType="HEADING_1"
)

Gotchas

1. Indices change after edits - Always re-fetch content after updating text
2. Indices are 1-based - First character is index 1, not 0
3. Newlines count - Each \n is one character in the index
4. Formatting is additive - Multiple calls can layer formatting
5. No markdown rendering - Symbols like # * | appear literally

Integration with Research Executor

The research-executor agent MUST use this skill for all Google Docs documentation at checkpoints.

Research Executor Checkpoint Structure:

• Experiment: {name} → HEADING_1
• Phase: {phase} → HEADING_2
• Timestamp → Italic normal text
• Sections (Hypothesis, Progress, Results, etc.) → HEADING_2
• Key findings/numbers → Bold

Workflow:

1. Write checkpoint content to Google Doc
2. Get document content with indices
3. Apply paragraph styles to headers
4. Bold key metrics and findings

Related Skills and Agents

• /gdrive-manager - Use BEFORE this skill to create documents in correct folder
• research-executor agent - Primary user for experiment documentation
• weekly-update-writer agent - Creates weekly updates
• research-slides-architect agent - Creates presentations
• /using-git-worktrees - Creates worktrees for experiments
• /finish-branch - Cleans up after PR is merged