delegate

Delegate bounded tasks to child Claude or Codex agents with monitoring and verification.

7|3|Updated Sep 3, 2026
One-click install
npx skills add https://github.com/gyroid-eth/orrery-telemetry --skill delegate-gyroid-eth
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: delegate
Source: https://github.com/gyroid-eth/orrery-telemetry/tree/main/skills/delegate
Command: npx skills add https://github.com/gyroid-eth/orrery-telemetry --skill delegate-gyroid-eth

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running multiple coding agents means manually copying output between them and checking each terminal to see what is happening. This Skill lets a parent agent hand off a bounded task to a child Claude or Codex agent, track its progress on a shared dashboard, and verify the result before reporting back. ## Core Features & Use Cases - Risk-aware task scoping: Profiles target files, exclusivity needs, and stuck points before spawning, and reserves shared file paths through ORRERY Mail to prevent collisions. - Managed child lifecycle: Pre-registers the child with a scientist name and token, spawns it in tmux via spawn_child.sh, annotates the dashboard with role and group, and monitors with risk-based check intervals. - Verified completion handling: Waits for the child's completion report with agentstack-await-reply, inspects changed artifacts, releases file reservations, and reuses or retires the child. - Use Case: Ask the agent to run a parallel code review in an isolated git worktree while you keep working; the child reports back by mail and the parent verifies the diff before merging. ## Quick Start Ask the agent to delegate a bounded task such as reviewing a module or fixing a flaky test to a child Claude or Codex agent, optionally specifying a model, effort level, or worktree isolation.

Frequently Asked Questions about delegate

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I delegate a task to a child Claude or Codex agent?

Invoke the delegate command with a task description, optionally adding flags for the working directory, Codex program, model, effort level, or worktree isolation. The parent agent registers the child, spawns it in tmux, and monitors it until a completion report arrives.

How do I run parallel agents without file conflicts?

Reserve the relevant file paths through ORRERY Mail before spawning each child, and give each child a bounded scope. For fully isolated edits, use worktree mode so each child works on its own branch in a separate git worktree.

Can I choose which model the child agent uses?

Yes. Claude children default to claude-opus-5 and accept opus, sonnet, or haiku shorthand. Codex children default to gpt-5.6-sol at xhigh effort and accept sol, terra, luna, or astra model shorthand plus an effort level.

How does the parent know when the child agent finished?

The child sends a completion report to the parent over ORRERY Mail at high importance, which arrives as a notification. The parent can also wait explicitly with the agentstack-await-reply command instead of polling the inbox.

Why does the monitor report exit code 10 when the child is still working?

Exit code 10 means a shell prompt pattern was detected, but the Claude Code REPL prompt uses the same character, so healthy children can trigger it between tool calls. Treat it as a signal to verify the inbox and pane, not as proof of completion.

When should I not delegate a task to a child agent?

Avoid delegating unbounded work without clear ownership or tasks where the required ORRERY Mail coordination tools are unavailable and the documented recovery path fails. Single-agent users gain little, since the value comes from multi-agent coordination.