Struct DelegateRecord 

pub struct DelegateRecord {
    pub sub_agent_id: String,
    pub target_agent_id: String,
    pub task: String,
    pub resolved_provider: String,
    pub resolved_model: String,
    pub usage: Usage,
    pub succeeded: bool,
    pub error: String,
    pub first_party: bool,
}

One __delegate_to call this turn dispatched (#872) — the forensic record of a worker sub-agent invocation.

Self::sub_agent_id is the identifier every forensic event for this delegation is tagged with — the control plane’s subagent_spawn, subagent_result, and subagent_model_call events all carry it, so a reader can join a worker’s spawn, its determinism inputs, and its result (and the visible tool_call/tool_result pair already in the transcript) by that one identifier.

Fields

sub_agent_id: String

The __delegate_to call’s provider-assigned tool-call id. Doubles as the sub-agent identifier (see the struct docs).

target_agent_id: String

The worker Agent resource name the model requested.

task: String

The self-contained task text handed to the worker (the model’s task argument; the optional context argument is not carried here — it rides on the worker’s own nested transcript, not this record).

resolved_provider: String

The worker’s resolved provider selector. Empty when the call was refused before a worker was resolved (a malformed call or an unmatched target).

resolved_model: String

The worker’s resolved model id. Empty under the same condition as Self::resolved_provider.

usage: Usage

Token usage the worker’s nested turn accumulated across its own provider calls. Zeroed when the call was refused before a worker ran.

succeeded: bool

true when the worker turn completed and produced an answer that became the __delegate_to call’s tool result; false on a malformed call, an unmatched target, a mid-stream provider failure, a worker that produced no text, or (#871) an answer that never conformed to result_schema after the one bounded retry.

error: String

Plain-language failure reason when Self::succeeded is false; empty on success.

first_party: bool

Whether this call’s result is first-party (untainted) content (#873). Defaults to true — every synthetic/error result run_delegate_call authors itself (a malformed call, an unmatched target, an invalid result_schema, or a worker turn that never ran) carries no worker content, so there is nothing to taint. For a call that actually dispatched a worker, this is explicitly recomputed from worker_ingested_untrusted_content over that worker’s own transcript: false the moment the worker touched a taint-source tool. The parent turn’s dispatch loop stamps this straight onto the __delegate_to call’s own tool-result Message in place of the static per-tool-name check every other tool result uses.

Trait Implementations

impl Clone for DelegateRecord

fn clone(&self) -> DelegateRecord

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for DelegateRecord

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for DelegateRecord

fn default() -> Self

first_party defaults to true (see the field doc) — every other field’s zero value already means “not yet resolved” (empty string, zero usage, not succeeded), so this is the one field a derived #[derive(Default)] would get backwards.

impl Eq for DelegateRecord

impl PartialEq for DelegateRecord

fn eq(&self, other: &DelegateRecord) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for DelegateRecord