Skills Compliance
Conduct risk in support conversations
Policy compliance and good conduct are different tests, and an operation can pass the first while failing the second. An agent who follows every documented step and leaves the customer worse off than they should be has complied and behaved badly.
npx rulebase-skills install cx-conduct-risk-monitoringWhen to use it
Reach for this when someone says any of these — they are the phrases the skill itself triggers on:
- “monitor for mis-selling”
- “conduct risk in our support conversations”
- “are agents pressuring customers”
- “customers being obstructed from cancelling”
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 risk categories
Each needs its own detection approach, and they have different severities.
Detection: behavioural signals beat keyword lists
Language-based detection finds the crude cases. The informative signals are structural, and they are computable without reading anything: A complaint expressed and not logged, Redress dispersion for like cases.
Sampling: risk-weighted, and honest about what that means
Conduct issues are rare, so uniform random sampling spends nearly all its budget on routine conversations.
Distinguish individual from systemic
The response differs completely, so the classification is the analytical work: Systemic, Individual.
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
