chore: update configuration and development tooling

- Add .prompts to .gitignore for local prompt management
- Refactor CLAUDE.md with clearer guidance on i18n, design system, and coding principles
- Add manifest-creator skill for automated manifest generation
- Add export-vendors.mjs script for vendor data export
- Enhance URL validation script with improved error handling and reporting

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: ericyangpan

.claude/skills/manifest-creator/SKILL.md

@@ -0,0 +1,276 @@
+---
+name: manifest-creator
+description: Create manifest files for AI coding tools by extracting information from their websites
+---
+
+# Manifest Creator Skill
+
+Automatically create manifest JSON files for AI coding tools, models, providers, and vendors by intelligently extracting information from their websites.
+
+## Overview
+
+This skill automates the creation of manifest files in the `manifests/` directory. It uses web scraping and AI to extract structured information from product websites, following the project's JSON schema requirements.
+
+## Usage
+
+To create a new manifest, provide:
+
+1. **Manifest type**: cli, extension, ide, model, provider, or vendor
+2. **Product name**: The filename to create (without .json extension)
+3. **Website URL**: The official website URL (becomes the `websiteUrl` field)
+
+**Example:**
+
+```
+Create a manifest for:
+- Type: cli
+- Name: copilot-cli
+- URL: https://github.com/github/copilot-cli
+```
+
+## How It Works
+
+The skill follows this workflow:
+
+### 1. Schema Validation
+
+First, it identifies the correct JSON schema based on the manifest type:
+
+- `cli` → `manifests/schemas/clis.schema.json`
+- `extension` → `manifests/schemas/extensions.schema.json`
+- `ide` → `manifests/schemas/ides.schema.json`
+- `model` → `manifests/schemas/models.schema.json`
+- `provider` → `manifests/schemas/providers.schema.json`
+- `vendor` → `manifests/schemas/vendors.schema.json`
+
+### 2. Web Content Extraction
+
+The skill uses the WebFetch tool to:
+
+- Visit the provided website URL
+- Extract text content and structure
+- Identify key information (description, version, pricing, etc.)
+- Follow relevant links to find additional details
+
+**Important Rules:**
+
+- The `description` field MUST be summarized from actual website content
+- Never generate descriptions - always extract from the source
+- If information isn't found on the main page, the skill will explore subpages
+
+### 3. Platform URL Discovery
+
+For models and providers, the skill will search for third-party platform URLs:
+
+- **HuggingFace**: Search for model/org pages on huggingface.co
+- **Artificial Analysis**: Search for performance benchmarks
+- **OpenRouter**: Search for model availability
+
+Each platform URL is discovered using the WebSearch tool.
+
+### 4. Manifest Generation
+
+The skill creates a complete JSON manifest file with:
+
+- All required fields from the schema
+- Properly formatted URLs (HTTPS only)
+- Validated structure
+- Appropriate null values for unavailable fields
+
+### 5. File Output
+
+The manifest is written to: `manifests/<type>/<name>.json`
+
+### 6. Validation
+
+After creating the manifest file, the skill **MUST** run validation:
+
+```bash
+node scripts/validate-manifests.mjs
+```
+
+This validation script will:
+- Check schema compliance
+- Verify all required fields are present
+- Validate field formats (URLs, enums, etc.)
+- Ensure filename matches the `id` field
+- Report any errors with detailed messages
+
+**If validation fails**, the errors must be fixed before completing the task.
+
+## Supported Manifest Types
+
+### CLI Tools
+
+**Schema**: Extends app.schema.json (product → vendor-entity → entity)
+
+**Key Fields**:
+- Basic: id, name, description, websiteUrl, docsUrl
+- Vendor: vendor
+- Product: latestVersion, githubUrl, license, pricing, resourceUrls, communityUrls, relatedProducts
+- App: platforms (macOS, Windows, Linux with install/launch commands)
+
+**Example**: `manifests/clis/claude-code.json`
+
+### Extensions
+
+**Schema**: Extends product.schema.json + supportedIdes
+
+**Key Fields**:
+- Same as CLI (except platforms)
+- Additional: supportedIdes (array with ideId, marketplaceUrl, installUri)
+
+**Example**: `manifests/extensions/claude-code.json`
+
+### IDEs
+
+**Schema**: Extends app.schema.json (same as CLI)
+
+**Key Fields**:
+- Identical to CLI structure
+- Focuses on IDE-specific features
+
+**Example**: `manifests/ides/cursor.json`
+
+### Models
+
+**Schema**: Extends vendor-entity.schema.json
+
+**Key Fields**:
+- Basic: id, name, description, websiteUrl, docsUrl, vendor
+- Model-specific: size, totalContext, maxOutput, tokenPricing (input, output, cache)
+- platformUrls: huggingface, artificialAnalysis, openrouter
+
+**Example**: `manifests/models/claude-sonnet-4-5.json`
+
+### Providers
+
+**Schema**: Extends vendor-entity.schema.json
+
+**Key Fields**:
+- Basic: id, name, description, websiteUrl, docsUrl, vendor
+- Provider-specific: type (foundation-model-provider | model-service-provider), applyKeyUrl
+- platformUrls, communityUrls
+
+**Example**: `manifests/providers/openrouter.json`
+
+### Vendors
+
+**Schema**: Extends entity.schema.json
+
+**Key Fields**:
+- Basic: id, name, description, websiteUrl, docsUrl
+- Additional: communityUrls
+
+**Example**: `manifests/vendors/anthropic.json`
+
+## Field Extraction Guidelines
+
+### Description Field
+
+**CRITICAL**: Never generate descriptions. Always extract from website content.
+
+- Must be concise (max 200 characters)
+- Summarize the product's key features
+- Use content from the homepage or about page
+- Keep original tone and terminology
+
+### URLs
+
+All URLs must:
+- Use HTTPS protocol
+- Be valid and accessible
+- Match schema patterns (e.g., GitHub URLs must start with `https://github.com/`)
+
+### Pricing
+
+For products (CLI, IDE, Extension):
+- Extract actual pricing tiers
+- Include: name, value (number or null), currency, per (billing period), category
+- Use null for custom/enterprise pricing
+
+For models:
+- Extract token pricing ($/M tokens)
+- Include: input, output, cache (null if not applicable)
+
+### Platform URLs
+
+Use WebSearch to find:
+- HuggingFace: Model or organization page
+- Artificial Analysis: Benchmark/comparison page
+- OpenRouter: Model availability page
+
+Set to null if not found.
+
+### Community URLs
+
+Search for official social media presence:
+- LinkedIn (company page)
+- Twitter/X
+- GitHub (organization)
+- YouTube (channel)
+- Discord (server invite)
+- Reddit (official subreddit)
+- Blog (official blog)
+
+Set to null if not found.
+
+## Workflow Example
+
+**User Request:**
+```
+Create a manifest for the Cursor IDE
+URL: https://cursor.sh
+```
+
+**Skill Actions:**
+
+1. Identify schema: `manifests/schemas/ides.schema.json`
+2. Fetch main page: Extract name, description, key features
+3. Search for documentation: Find docs URL
+4. Search for GitHub: Find repository URL
+5. Extract version: Check for latest release
+6. Find pricing: Navigate to pricing page
+7. Gather community URLs: Search for social media links
+8. Generate manifest: Create `manifests/ides/cursor.json`
+9. **Run validation**: `node scripts/validate-manifests.mjs`
+10. Fix any validation errors if found
+
+## File Structure
+
+```
+.claude/skills/manifest-creator/
+├── SKILL.md              # This documentation
+└── scripts/
+    └── create.mjs        # Main creation script
+```
+
+## Important Notes
+
+1. **Never Auto-Generate Content**: Always extract from actual website sources
+2. **Verify URLs**: All URLs should be tested for accessibility
+3. **Follow Schema**: Strictly adhere to the JSON schema for each type
+4. **Use Null Appropriately**: Set unavailable fields to null (not empty string)
+5. **ID Format**: Always use lowercase-with-hyphens (e.g., `claude-code`, not `Claude Code`)
+6. **Verified Field**: Default to `false` unless explicitly verified by maintainers
+7. **ALWAYS Validate**: Run `node scripts/validate-manifests.mjs` after creating any manifest file
+
+## Error Handling
+
+If the skill cannot find required information:
+- Ask the user for manual input
+- Suggest alternative URLs to search
+- Provide partial manifest for user completion
+- Log missing fields clearly
+
+## Contributing
+
+When improving this skill:
+- Update schema mappings if new manifest types are added
+- Enhance web scraping patterns for better extraction
+- Add more platform URL sources
+- Improve description summarization
+
+## License
+
+This skill is part of the AI Coding Stack project and follows the project's license.