Skip to main content

Sort Tool Results by Trust Treatment

Classify tool results as safe to use, needs validation, needs human review, or must be blocked.

agent-tool-design5 min read

Place each tool result into the treatment it should receive before the model uses it. Pass as structured context Validate or sanitize first Human review required Block or stop Trusted CRM API row matching output schema and freshness window Warehouse count from internal API with required timestamp and ID Vendor web page with useful claims and marketing copy Customer email containing instructions to ignore company policy Refund eligible by amount but blocked by region policy Two possible customer records match the same email address Tool returns RATE_LIMIT with no business data Protocol error says the requested tool name is unknown

Read the full lesson

Sign up free: one personalized lesson every day, matched to your role and goals.

← Back to library

Contact us