Chrome Update Digest

Process Chrome release notes into structured, multilingual digests for 23 focus areas.

Design Principles

Stateless Processing:

• No version history tracking - each run is independent
• Only processes the specified version (e.g., "Process Chrome 143")
• Does not track what versions have been processed before

Self-contained & Portable:

• Vendored dependencies - no external package dependencies
• Bundled prompts for AI digest generation
• No dependencies on repo files outside the skill directory
• Works independently after MCP separation

Dual-source Release Notes:

• Chrome release notes (required): chrome-{version}-{channel}.md
• WebGPU release notes (optional): webgpu-{version}.md
• WebGPU enhances the graphics-webgpu area when available but is not required

Agent Delegation:

• Skill orchestrates the workflow (download → extract → generate navigation)
• AI digest generation delegated to agents with bundled prompts
• Agents process all 23 areas in parallel for efficiency

Quick Start

Simple usage (smart defaults):

Process Chrome 143

Uses: stable channel, bilingual (en+zh), all 23 areas

Custom usage:

Process Chrome 143 beta in English for CSS and WebAPI

Core Features:

• ✓ Extracts structured data from release notes into 23+ focus areas
• ✓ Generates AI-powered bilingual digests (English + Chinese)
• ✓ Creates GitHub Pages navigation (dual structure: by-version & by-area)

Quick Reference (For Claude Agent)

Workflow Overview

When user says "Process Chrome 143":

# Step 1: Extract areas (automatic download if missing)
uv run python .claude/skills/chrome-update-digest/scripts/process_chrome.py --version 143

# Step 2: Get area list
uv run python .claude/skills/chrome-update-digest/scripts/process_chrome.py --version 143 --list-areas
# Output: 23 areas (css, webapi, graphics-webgpu, etc.)

# Step 3: Generate digests (spawn agents in parallel)
# For each area:
#   - English: Read prompts/webplatform-prompt-en.md → Generate EN digest
#   - Chinese: Read prompts/webplatform-translation-prompt-zh.md → Generate ZH digest

# Step 4: Done! Navigation auto-generated in step 1

Commands Cheat Sheet

Agent Integration

Generate English Digest:

1. Read: prompts/webplatform-prompt-en.md
2. Replace: [AREA] → area name, [version] → 143, [channel] → stable
3. Input: upstream_docs/.../areas/{area}/chrome-143-stable.yml
4. Output: digest_markdown/webplatform/{area}/chrome-143-stable-en.md

Generate Chinese Translation:

1. Read: prompts/webplatform-translation-prompt-zh.md
2. Replace: [AREA], [version], [channel] with actual values
3. Input: English digest from previous step
4. Output: digest_markdown/webplatform/{area}/chrome-143-stable-zh.md

Process all 23 areas in parallel for efficiency.

Workflow

The skill executes these steps automatically:

1. Validate: Check version format and input files exist
2. Clean Pipeline: Extract area-specific YAML and markdown from raw release notes
3. Generate AI Digests: Use agent with bundled prompts to create bilingual summaries

- English digest: Use agent with prompts/webplatform-prompt-en.md - Chinese translation: Use agent with prompts/webplatform-translation-prompt-zh.md

1. Generate Navigation: Create GitHub Pages navigation structure (by-version & by-area)
2. Output: Save digest files and navigation to respective directories

Processing Options

Channel: stable (default) or beta Output: Area-specific markdown + YAML files + GitHub Pages navigation

Reference Documentation

• [Focus Areas](references/focus_areas.md) - 23 focus areas definition
• [Implementation Guide](references/implementation_guide.md) - Detailed step-by-step implementation
• [Error Handling](references/error_handling.md) - Comprehensive error scenarios and solutions
• [Testing & Validation](references/testing_validation.md) - Post-processing validation checklist
• [Common Issues](references/common_issues.md) - Troubleshooting guide
• [Prompts Usage](prompts/README.md) - Prompt template documentation

Examples

Example 1: Simple request with defaults

User: "Process Chrome 143"
Assistant:
  Parsing request: version=143, channel=stable
  Validating inputs... ✓

  🚀 Processing Chrome 143 (stable)

  [1/4] Running clean data pipeline...
    ✓ Extracted 23 focus areas
    ✓ Generated YAML and markdown files

  [2/4] Generating AI digests for 23 areas...
    Launching agents in parallel for English digests...
    ✓ css (English)
    ✓ webapi (English)
    ✓ graphics-webgpu (English)
    ... (20 more areas)

    Launching agents in parallel for Chinese translations...
    ✓ css (Chinese)
    ✓ webapi (Chinese)
    ✓ graphics-webgpu (Chinese)
    ... (20 more areas)

  [3/4] Generating GitHub Pages navigation...
    ✓ Created version-centric pages
    ✓ Created area-centric pages

  [4/4] Validation...
    ✓ All digest files generated
    ✓ Navigation includes Chrome 143

  ✅ Processing complete!
     Processed areas: upstream_docs/processed_releasenotes/processed_forwebplatform/areas/
     AI Digests: digest_markdown/webplatform/
     Navigation: digest_markdown/areas/ and digest_markdown/versions/

  💡 Next steps:
     1. Review generated digests in digest_markdown/webplatform/
     2. Commit changes to trigger GitHub Pages deployment

Example 2: Check latest version

User: "What's the latest Chrome version?"
Assistant:
  Checking latest Chrome versions...

  Latest Chrome versions:
    Stable: 143 (2026-01-15)
    Beta:   144 (2026-01-10)

  To process: "Process Chrome 143" or "Process Chrome 144 beta"

Example 3: Beta channel

User: "Process Chrome 143 beta"
Assistant:
  Parsing request: version=143, channel=beta
  Validating inputs... ✓

  [Processing output similar to Example 1, but for beta channel...]

Example 4: Error handling

User: "Process Chrome 999"
Assistant:
  Parsing request: version=999, channel=stable, language=bilingual, areas=all
  Validating inputs... ✗

  Error: Chrome 999 release notes not found.
  Expected: upstream_docs/release_notes/WebPlatform/chrome-999.md

  Solutions:
  1. Check if version 999 is released yet (latest is 143)
  2. Try a different version: "Process Chrome 143"
  3. Download release notes from: https://developer.chrome.com/release-notes/

Technical Notes

• Stateless Design: No version history tracking - each run is independent
• Self-contained: Vendored dependencies, no external package requirements
• Dual-source Release Notes:

Chrome WebPlatform (required): Main release notes WebGPU (optional): Enhances graphics-webgpu area when available

• The skill orchestrates processing workflow via process_chrome.py
• Uses uv run to ensure correct environment and dependencies
• Progress is streamed in real-time
• AI Digest Generation:

Uses Task tool with general-purpose agents Processes all areas in parallel for efficiency Bundled prompts: English digest + Chinese translation (see [prompts/README.md](prompts/README.md)) Each agent receives YAML content and produces markdown digest

• Helper Functions:

--check-latest: Query latest Chrome stable/beta versions (stateless) --list-areas: List areas extracted for a specific version * getareasto_process(): Python helper for agent integration

• Outputs:

Processed files: upstreamdocs/processedreleasenotes/processedforwebplatform/areas/ AI Digests: digestmarkdown/webplatform/{area}/chrome-{version}-{channel}-{lang}.md * Navigation: digestmarkdown/areas/ and digestmarkdown/versions/

• Supports both stable and beta channels
• Handles 23+ predefined focus areas (see config/focus_areas.yaml)
• Bundles configuration and prompts for portability
• Integrates with GitHub Actions for automatic deployment

Prompt Details

The skill includes two bundled prompts for AI digest generation:

1. English Digest Prompt ([prompts/webplatform-prompt-en.md](prompts/webplatform-prompt-en.md))

- Input: Area-specific YAML file - Output: Structured English digest with feature descriptions - Format: Markdown with consistent structure

1. Chinese Translation Prompt ([prompts/webplatform-translation-prompt-zh.md](prompts/webplatform-translation-prompt-zh.md))

- Input: English digest content - Output: Chinese translation maintaining technical accuracy - Format: Preserves markdown structure and links