Skills Data and integration
Canonical conversation schema
One schema for support conversations across every platform, so a metric is written once and runs anywhere. Includes a validator that catches the export problems which look fine and produce wrong answers.
npx rulebase-skills install cx-conversation-schemaWhen to use it
Reach for this when someone says any of these — they are the phrases the skill itself triggers on:
- “normalise our support data”
- “unify conversations from multiple helpdesks”
- “validate our conversation export”
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.
Why normalise at all
Every helpdesk models the same thing differently, and the differences are exactly where analyses break: Zendesk has 7 statuses, Freshdesk has 4 numeric codes plus custom ones, "Internal note" is public.
The schema
Two newline-delimited JSON files. JSONL because exports stream, append safely across resumed runs, and never need to fit in memory.
Four rules that carry all the weight
1. source_id is always a string. Zendesk and Freshdesk use integers, Intercom uses numeric strings, Five9 uses opaque call ids. Left as native types they compare unequal across systems, and large integer ids lose precision in JavaScript and some warehouse loaders. Stringify at export.
Validate before you analyse
It checks structure — required fields, enum vocabulary, ISO timestamps, csat in range, duplicate ids — and the semantic problems that matter more.
Deriving analysis inputs
The canonical export is the base layer. Most analyses need a derived view.
Cross-platform identity
The hardest part of a multi-platform model, and the one that quietly invalidates analyses.
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
