Skip to content

Agents and automation

This chapter follows work that leaves the single foreground model turn: built-in agents, custom agents, skills, subagents, background and multi-turn tasks, autopilot/no-ask-user modes, fleet orchestration, scheduled prompts, and command queues.

Read this chapter when the question is: how does the runtime delegate, parallelize, continue, or schedule work beyond one assistant response?

Source-anchor policy

This page is a chapter guide. Linked implementation pages carry concrete app.js anchors.

Semantic alias	Minified anchor	Scope
Agents/automation chapter	N/A — navigation page	Groups task orchestration, built-in/custom agents, skills, autopilot/no-ask-user, fleet, and scheduled prompts.
Agent implementation pages	See linked source-anchor tables	Concrete bundle anchors live in the destination pages.

Automation map

flowchart TD
    Main[Main session/model] --> TaskTool[task tool]
    TaskTool --> Registry[TaskRegistry]
    Registry --> Builtins[Built-in agents]
    Registry --> Custom[Custom agents]
    Registry --> MCP[MCP tasks]
    Custom --> Skills[Skills / AGENTS.md / allowed-tools]
    Main --> Autopilot[Autopilot and no-ask-user]
    Main --> Fleet[Fleet mode]
    Main --> Schedules[Scheduled prompts]
    Fleet --> Registry
    Schedules --> Queue[Command queue]
    Registry --> Results[Result handoff / lifecycle events]

    click Registry "./agent-task-orchestration/" "Open agent task orchestration"
    click Builtins "./built-in-agents/" "Open built-in agents"
    click Custom "./custom-agents-and-skills-packaging/" "Open custom agents and skills"
    click Autopilot "./autopilot-and-no-ask-user/" "Open autopilot"
    click Fleet "./fleet-mode/" "Open fleet mode"
    click Schedules "./scheduled-prompts-and-command-queue/" "Open scheduled prompts"

Primary reading order

Order	Page	Agent/automation question answered
1	Agent and task orchestration	How do the task tool, TaskRegistry, main/subagent/custom-agent collaboration, hooks, MCP tasks, research, and fleet share lifecycle state?
2	Built-in agents	Which built-in agent types exist, how are YAML/runtime-defined prompts selected, and what slash-command entry points use them?
3	Custom agents and skills packaging	How do AGENTS.md, SKILL.md, provided/remote/plugin agents, skill directories, invocation, allowed-tools, and enable/disable events work?
4	Autopilot and no-ask-user flags	How do `--autopilot`, `--no-ask-user`, continuation, `task_complete`, `ask_user`, and permission boundaries differ?
5	Fleet mode implementation	How do `/fleet`, `session.fleet.start`, SQL todo coordination, dependencies, and parallel subagents interact?
6	Scheduled prompts and command queue	How do `/every`, `/after`, ScheduleRegistry replay, queue integration, and ephemeral command dispatch work?

Handoffs

Agent prompts and skills feed the Context and model loop.
Agent-specific tools and approval boundaries use Tools, integrations, and security.
Background/multi-turn tasks persist through Sessions, persistence, and remote.
Hosted job constraints and validation toggles are covered by Hosted agent ops.

Created and maintained by Yingting Huang.