Skills Quality assurance
Auditing a support bot for harm
Containment and deflection metrics measure whether a bot reduces volume. They say nothing about whether it is causing harm, and a bot can score well on both while doing all four of the things below.
Install
npx rulebase-skills install cx-bot-safety-auditWhen to use it
Reach for this when someone says any of these — they are the phrases the skill itself triggers on:
- “are people trying to jailbreak our bot”
- “show me attempts to manipulate the assistant”
- “is our bot giving wrong answers”
- “customers stuck with the bot and never got a human”
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 four risk classes
Audit them separately. They have different detection methods, different severities and different owners.
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
