Skip to content

evaluatePromptPolicies()

PromptScript API


evaluatePromptPolicies(policies, subjects, context?, scope?): PolicyViolation[]

Defined in: validator/src/policy/prompt-evaluator.ts:248

Evaluate the prompt policy kinds on prompts.

Every policy matches on policy risk: the higher of the effective risk (unset counts as low) and the floor of every risk-floor policy, whatever its severity. required-approvals, required-evals and max-experiment-weight are not evaluated here.

Parameters

policies

readonly PolicyDefinition[]

Parsed policy definitions

subjects

readonly PolicySubject[]

Prompts as policies see them

context?

PolicyContext = {}

Model catalog settings and the blocked-pattern setup of PS005

scope?

PromptPolicyScope

Risk and selection to use for every subject instead of its own

Returns

PolicyViolation[]

Violations, each at the location of its prompt