Website Flow Monitor#

Take a user-provided website URL and build a practical monitoring plan.

Workflow#

1. Normalize input URL (ensure https://).
2. Discover likely flows by scanning:
   • homepage
   • pricing
   • docs/api
   • blog/changelog
   • downloads/signup/login/dashboard if present
   • robots/sitemaps
3. Categorize checks by impact:
   • Revenue: pricing, checkout, billing links, purchase CTAs
   • Onboarding: signup/login/download/get-started flows
   • Product/API: docs and API endpoints (including expected auth failures like 401)
   • Trust: privacy, terms, contact/support
   • SEO/Discovery: robots.txt and sitemaps
   • Content: blog posts, changelog entries
4. Return a proposed monitoring set with:
   • URL
   • expected behavior (200, redirect, 401/403, etc.)
   • impact category
5. Ask for confirmation before scheduling cron.
6. After confirmation, create/update a cron job with:
   • sessionTarget: isolated
   • concise agentTurn task that checks all agreed flows
   • delivery mode announce

Rules#

• Do not schedule cron until user confirms the proposed flow list and frequency.
• Prefer actionable summaries: failures first, then healthy checks.
• Flag as ALERT for 4xx/5xx (except expected 401/403 on auth-required API checks), redirect loops, or missing critical links.
• If the user requests changes (skip/add endpoint), update the existing cron job instead of creating duplicates.

Discovery helper#

Use this script for quick link discovery:

bash

python3 {baseDir}/scripts/discover_flows.py --url https://example.com

It returns JSON with discovered internal/external links and suggested flow candidates.

Cron prompt template (adapt per site)#

Use this shape in payload.message:

• mention this is a recurring reminder
• list exact URLs and expected behavior
• require output format: OK / WARN / ALERT
• require impact tags: revenue/onboarding/product/trust/seo/support
• include "if all healthy, short OK summary" instruction

Keep it short and deterministic.