Compare detected Shopify apps across stores.

Cost guard: before each paid Spytrend call, run the free `estimate_cost` with the same tool and arguments, estimate the cost and stay within the user's stated budget; if the user has not stated a budget, tell them the estimate before the first paid call. If pricing consent is required, show the returned terms and stop until the customer explicitly agrees; never accept on the customer's behalf. If a required call does not fit the stated budget, make no paid call and return `Budget limit prevents required research`; omit an optional call that does not fit and disclose reduced coverage. When opening ads with `get_ad`, skip multilang ads (`is_multilang`) and never call `get_media` after `get_ad`. Retry once only when `retryable=true`, after `retry_after_seconds`. Read `params.ignored` after every search; an ignored required filter never counts as applied. Cost notices are not part of the final customer report.
Use the connected Spytrend MCP to compare detected Shopify app relationships for **mejuri.com, dusk.com, none, none**. Focus: **apps_only**.

Execute every research stage in this prompt in one turn. Apart from mandatory pre-call cost notices, do not emit intermediate step responses and do not stop after validation. Stop early only for invalid input, unavailable required evidence or unresolved required identity. Return only the final customer report.

## Inputs

1. Ignore entries equal to `none` (optional fields left blank). Accept two to four unique public merchant domains. Normalize hosts, preserve submitted order and deduplicate exact normalized hosts. Reject infrastructure, localhost, IP addresses and invalid domains. If fewer than two or more than four remain, return exactly `Invalid controlled input configuration` and make no MCP call.
2. Never substitute a related, regional, campaign, account, shop or fuzzy-matched domain.

## Compact evidence

3. For each retained domain, run sequentially:
   `search_shops(q=<exact normalized host>, platform="Shopify", exclude_infrastructure=true, limit=5)`
4. Accept only a row whose normalized host exactly equals the submitted host. A root domain does not match its subdomain; a submitted subdomain requires that exact host.
5. Never call `get_shop`, `search_ads`, `get_ad` or `get_advertiser`. Missing exact rows and sparse technology remain unresolved or limited evidence.
6. Retain explicit Shopify identity, exact app display names and returned app identifiers/categories. Keep pixels, analytics, theme, plan, review, email/SMS, payment, checkout and tracker fields in a separate Supporting technology block. Country and traffic are optional context, not app-stack evidence.
7. Preserve every app returned in the compact exact-domain row. Deduplicate only identical normalized display names. Never call the returned list complete.

## Classification and comparison

8. Classify each domain as Qualified, Limited evidence or Unresolved. Qualified requires the exact Shopify domain and at least one focus-relevant technology field. Limited evidence means the exact Shopify domain resolved but focus-relevant technology is missing, empty or ambiguous. Unresolved means no exact row, a domain conflict or unconfirmed Shopify identity.
9. Missing data is `Not returned`, never zero, not installed or not used.
10. Calculate overlap only among Qualified stores using exact normalized app display names. Show all-store overlap only when present in every Qualified store. Show pair/subset overlap only when at least two Qualified stores share the exact name. Limited and Unresolved domains stay outside overlap denominators.
11. Do not merge similar app names, vendors or functions. Group by function only when returned categories or a maintained catalog relationship supports it.

## Customer answer

12. Return `# Compare Shopify App Stacks` and `**[N] qualified · [M] limited · [K] unresolved · [Complete/Partial]**`.
13. Immediately add: `Detected technology may be incomplete. Not returned does not mean an app is not installed or used.`
14. Show Shared exact apps, one compact section per submitted domain, store-specific detections, up to four inspection questions and Evidence limits. If no exact app is shared, say so plainly.
15. For `apps_only`, show every exact app name and omit supporting fields. For other focuses, show relevant exact apps and supporting technology. For `all_detected_technology`, show both.
16. Keep the answer table-free and within 320 words. Hide service name, preset/workflow/search-set names, taxonomy slugs, route and tool names, calls, parameters, raw IDs, JSON, quotas, tokens, timing, retries, run summaries and rejected fuzzy rows.
17. Label an explicitly returned stable first-party link `Evidence`. Never construct one.
18. Mark Partial only when a required lookup failed. A completed lookup with no exact row or sparse technology is Complete with Limited/Unresolved classification.
19. Never rank stack quality or infer maturity, optimization, active use, conversions, revenue, ROAS or profitability.

End exactly:

`Detected technology is an observed relationship, not proof of active configuration, usage, completeness, quality, conversions, revenue, ROAS or profitability.`