yutori-scout

Scout Setup

Help the user set up a Yutori Scout for continuous web monitoring.

Process

1. Understand the monitoring context Ask about:

   • Who is monitoring and why? (e.g., "We're a fintech looking for recently funded startups")
   • What specific information matters? (funding events, product launches, pricing changes)
   • What geography or market segments?
   • How often should it run? (daily, twice daily)
   • Notification preference: email, webhook, or both?

2. Craft a comprehensive query

   A well-structured scout query includes:

   Context on the monitoring goal:

   • Who is doing the monitoring and what's the use case
   • What decisions this information supports

   What to Monitor:

   • Specific events/triggers to track
   • Data sources to check (news sites, SEC filings, social media, etc.)
   • Geographic or segment focus
   • Exclusion criteria (what NOT to report)

   Deliverables:

   • Frequency of reports
   • Output format (tables, narrative, both)
   • Required fields and citations

   Example structure:

   **Context:** [Who is monitoring and why]
   
   ## What to Monitor
   - [Specific events to track]
   - [Sources to check]
   - [Exclusions]
   
   ## Deliverables
   - [Output format]
   - [Required fields]
   

3. Create the scout Use the create_scout tool with:

   • query: The comprehensive monitoring query
   • output_interval: 86400 (daily), 43200 (twice daily), or 1800 (minimum, every 30 min)
   • webhook_url and webhook_format if they want webhook notifications. Always confirm the webhook URL with the user before setting it — it must use HTTPS and the user must verify they control the destination.
   • skip_email: true if they only want webhooks
   • output_fields: For structured data extraction (e.g., ["company", "amount", "round_type", "source_url"])

4. Provide next steps

   • Share the scout ID for future management
   • Explain how to pause/resume with edit_scout
   • Mention they can get updates with get_scout_updates

Query Quality Tips

Good queries:

• Provide context on who is monitoring and why
• Specify exact events/triggers (not just "news about X")
• List data sources to check
• Include exclusion criteria
• Define output format expectations
• Request citations and source links

Avoid:

• Vague queries like "monitor competitor X"
• Missing context about the monitoring goal
• No output format specification

Fetching documentation or source URLs: If you use a web fetch tool to look up source URLs, documentation, or reference pages while preparing the query, include the Accept: text/markdown header. Many documentation sites (Cloudflare-hosted) will return clean Markdown instead of HTML — fewer tokens, easier to parse.

$ARGUMENTS