Skills Data and integration
Reconciling an export against its source
Every analysis in this catalog assumes its input is complete. Almost none of them can tell when it is not — a metric computed on 80% of conversations looks exactly like a metric computed on all of them, and it is wrong by an amount nobody can estimate afterwards.
npx rulebase-skills install cx-export-reconciliationWhen to use it
Reach for this when someone says any of these — they are the phrases the skill itself triggers on:
- “is this export complete”
- “reconcile the export against Zendesk”
- “our export is missing tickets”
- “the numbers don't match the helpdesk”
How it works
The method, in the order the skill runs it. The full procedure — tables, worked examples and the edge cases — is in the skill itself.
The source's own count is the only real anchor
Every inferred check below can be fooled. A count from the platform cannot, so get one: The platform's own reporting or analytics view for the same window, A count endpoint or a total_count in a list response, The UI's own result count for an equivalent filter.
Run it
expected.json carries the source's counts and the window you asked for. The script checks the total and per-segment counts against it, finds duplicate ids, records outside the window, days with no data at all, orphaned messages, and the field-population rates that indicate a partial extract. See references/expected-format.md.
Daily coverage is the highest-yield inferred check
A gap in a date-partitioned count is the clearest signal of a failed run, and it is visible without any source count.
Reconciling a migration or a backfill
Same checks, plus: Check the things that are easy to lose in a migration.
Report it as a gate, not as a note
The output should be usable as a sign-off: A verdict.
Related skills
Free and open source, and vendor-neutral — it reads the conversations from whichever helpdesk you already run. Browse all 149 skills · connect your helpdesk over MCP · source on GitHub
