Build and maintain an AI-driven news curation platform with automated elaboration, commentary capture, and multi-platform publishing capabilities using AWS serverless architecture.
You are working on an AI-driven news curation and commentary platform that aggregates content from RSS feeds and APIs, provides AI-powered elaborations with strict citation requirements, captures personal commentary through Q&A prompts, and generates publishable drafts for LinkedIn and Substack.
**Stack Overview:**
**Project Structure:**
```
/backend - Lambda functions and API server
/frontend - Remix/Next.js web application
/shared - Shared TypeScript types and utilities
/docs - Documentation and PRD
/tests - Test files (Vitest)
/database - Migrations and seeds
```
Before EVERY commit, you MUST run all validation steps:
1. **Build**: `npm run build` - Verify compilation succeeds
2. **Lint**: `npm run lint` - Check code style
3. **Type Check**: `npm run type-check` - Verify TypeScript types
4. **Tests**: `npm run test` - Run full test suite
**NEVER commit failing code.** Fix all errors before proceeding.
1. **Make Changes**: Implement features, write tests, update documentation
2. **Validate Code**: Run ALL validation commands listed above
3. **Review Changes**:
- `git status` - See modified files
- `git diff` - Review line-by-line changes
4. **Stage Files**: `git add <files>` - Stage only relevant changes
5. **Commit Changes**: Use clean, professional commit messages
```
<type>: <description>
Types: feat, fix, test, chore, refactor, docs
Example: test: add elaboration citation validation tests
```
**CRITICAL RULES:**
```
type/personid-linearid-description
Example: feature/macs-MAC-121-ticket-prd
```
```bash
npm install # Install dependencies
npm run dev # Start dev server (backend:3001, frontend:3000)
npm run build # Build project
npm run clean # Clean build artifacts
npm run test # Run tests
npm run test:watch # Run tests in watch mode
npm run lint # Lint code
npm run type-check # TypeScript type checking
npm run db:migrate # Run migrations
npm run db:seed # Seed database
```
**Data Sources:**
**Feed Card Requirements:**
**Elaboration Requirements:**
**LLM Strategy:**
**Performance Targets:**
Generate 3-5 POV prompts per article:
**Requirements:**
**Platform-Specific Templates:**
**Draft Features:**
**OAuth Integration:**
**Success Criteria:**
**Scheduler:**
**Roundup Generation:**
**Users**: User profiles, preferences, topics, keywords, regions, sources
**Sources**: RSS feeds and API sources with trust scores
**Items**: Normalized articles with dedupe clustering
**UserItemState**: User actions (viewed, saved, dismissed, interesting)
**Elaborations**: AI summaries with citations and takeaways
**QnAResponses**: User commentary per item
**Drafts**: Generated content with version history
**Roundups**: Scheduled content compilations
**Tokens**: Encrypted OAuth credentials
`items_idx` - For ranking, search, and relevance scoring
**Required API Keys:**
**AWS Services:**
Follow the detailed implementation plan in `docs/milestone-1-implementation-plan.md` for step-by-step guidance on building each feature.
1. **Serverless-first**: All backend logic must run on AWS Lambda
2. **Strict attribution**: Every AI elaboration MUST include verified citations
3. **TypeScript end-to-end**: No JavaScript files except config
4. **Test coverage**: All features require Vitest tests before merge
5. **Security**: OAuth tokens encrypted, KMS for sensitive data, no hardcoded credentials
Leave a review
No reviews yet. Be the first to review this skill!
# Download SKILL.md from killerskills.ai/api/skills/ai-news-aggregator-system/raw