Research Skill

Overview

This skill orchestrates AI-powered research workflows using Perplexity API integration. It handles daily monitoring, topic discovery, comprehensive research reports, and content project kickoffs.

For usage examples and workflows, see CLAUDE.md.

Operations

daily - Daily news monitoring

• Project-based: Monitor configured sources (Reddit, HN, RSS) for project
• Standalone: Generate digest based on saved topics
• Entry: research-skill daily [project-name]

discover - Topic discovery and exploration

• Broad topic → Extract subtopics → Deep dive on selected topics
• Entry: research-skill discover "<topic>"

query - Direct research query

• Execute specific research question with appropriate recency
• Entry: research-skill query "<question>"

browse - Browse research history

• Review past research, organized by date
• Entry: research-skill browse

kickoff - Kickoff content project from research

• Use research file as foundation for new content project
• Entry: research-skill kickoff <research-file> <project-name>

setup-monitoring - Configure project-based monitoring

• Interactive setup for monitoring sources (subreddits, HN, RSS)
• Creates monitoring-config.yaml in project
• Entry: research-skill setup-monitoring <project-name>

Technical Details

CLI Integration

Uses kurt research CLI commands:

• kurt research search "<query>" --recency <hour|day|week|month> --save
• kurt research monitor <project-dir>
• kurt research list
• kurt research get <slug>

Configuration

Location: .kurt/research-config.json (gitignored)

Models:

• sonar-reasoning - Best for comprehensive research (default)
• sonar - Faster, good for quick queries
• sonar-pro - Most powerful, higher cost

Recency filters: hour, day, week, month

Output Format

Research files: sources/research/YYYY-MM-DD-<query>.md

Contains:

• YAML frontmatter (query, citations, metadata)
• Comprehensive answer with inline citations
• Full source list

Monitoring signals: projects/<name>/research/signals/YYYY-MM-DD-signal.md

Conversational Mode

When invoked from gather-sources subskill:

1. User mentions research need
2. Ask clarifying questions to refine query
3. Show proposed query
4. Get user approval before API call
5. Execute and save results

Key Differences: Research vs Documents

Research files:

• Ephemeral (news, trends, time-sensitive)
• Saved as markdown in sources/research/ or projects/<name>/research/
• NOT imported to Kurt database
• Can be referenced as sources in content creation
• Browse with kurt research list (not kurt content list)

CMS/Web content:

• Documentation, articles, reference material
• Stable content that updates over time
• Imported to Kurt DB for indexing and clustering
• Browse with kurt content list

Integration with Other Skills

Called by:

• gather-sources subskill - Routes research requests here with conversational mode
• Direct invocation for daily workflows

Delegates to:

• kurt research CLI - For API interaction with Perplexity
• content-writing-skill - For creating content from research (via kickoff operation)

Works with:

• project-management-skill - Research files can be sources in projects
• cms-interaction-skill - Research → Content → CMS publish workflow

Key Principles

1. Conversational refinement - Ask clarifying questions before expensive API calls
2. Research ≠ Documents - Separate from Kurt DB, saved as reference files
3. Project-based monitoring - Research lives with project for clear lineage
4. Citation tracking - All research includes full source attribution
5. Recency awareness - Use appropriate time filters for different query types

For workflows and usage examples, see CLAUDE.md. For monitoring configuration details, see .kurt/README.md.