Module agents 

Agent specs and prompt packet rules.

Operator framing

Use this page when you want to understand what an AI workflow is allowed to see, draft, or summarize before a person reviews it. It is for operators, managers, and implementers checking the safety gate between helpful automation and live resort actions.

The next step is to read baseline_agent_specs for the built-in workflow list, then open AgentPromptPacket to see the evidence bundle each run receives. The Rust API details below remain the implementer rules; the framing here explains how to interpret those rules operationally.

This module is the app-layer rules for safe workflow automation. Agent specs describe narrow read/draft capabilities for resort workflows, and prompt packets carry the source event, typed workflow input, policy language, and expected output schema an agent may use to prepare a draft or evidence bundle. They are not live authority to mutate bookings, customer messages, schedules, deposits, incident records, or policy decisions.

Specs make the app gate visible to the runtime: allowed tools are narrow read/draft surfaces, forbidden actions name unsafe authority, and default review gates remain deterministic app policy.

use app::agents;

let specs = agents::baseline_agent_specs();
let manager_brief = specs
    .iter()
    .find(|spec| spec.name.clone().into_inner() == "manager-daily-brief")
    .expect("baseline manager brief spec exists");

assert!(manager_brief
    .allowed_tools
    .iter()
    .any(|tool| tool.clone().into_inner() == "reservation-read"));
assert!(manager_brief
    .forbidden_actions
    .iter()
    .any(|action| action.clone().into_inner() == "change schedule"));
assert!(manager_brief
    .forbidden_actions
    .iter()
    .any(|action| action.clone().into_inner() == "send customer message without approval"));
assert!(!manager_brief.default_review_gates.is_empty());

Structs

AgentPromptPacket

Safe prompt-and-evidence packet exchanged with an automation agent.

AgentPromptPacketBuilder

Use builder syntax to set the inputs and finish with build().

OutputSchemaName

PolicyInstruction

Traits

WorkflowAgent

Rules implemented by app workflow agents that prepare safe prompt packets.

Functions

baseline_agent_specs

Returns the baseline app agent specifications for pet-resort automation.

Type Aliases

AgentSpec

App-facing alias for the domain agent specification used by workflow automation.