vigiles API reference
    Preparing search index...

    Interface ModelRequest

    One model request the scripted mock received, flattened to text for assertions — the seam that lets a harness test prove what reached the model (a SessionStart hook's injected context, a slash command's expansion), not just that a hook fired.

    interface ModelRequest {
        messages: readonly { role: string; text: string }[];
        sideChannel?: boolean;
        system: string;
    }
    Index

    Properties

    messages sideChannel? system

    Properties

    messages: readonly { role: string; text: string }[]

    The conversation messages, each flattened to { role, text }.

    sideChannel?: boolean

    true when the agent CLI made this call for its OWN bookkeeping rather than to advance the agent loop — a SIDE-CHANNEL call.

    MEASURED on Claude Code 2.1.228: a post_turn_summary classifier (it decides whether to push the user a phone notification) runs on every turn, on the user's own credential, and it declares NO tools. The mock does not let such a request consume a script entry — see isMainLoopRequest in mock-model.ts.

    system: string

    The system prompt, flattened to text (string or text-block array).

    Settings

    Member Visibility

    On This Page

    Properties