Orchestrate delegated research or implementation with Claude or Codex agents. Default to Claude in Claude Code and Codex in every other harness. Follow an explicit user choice of agent or model. Plan and launch without a separate plan approval step.
Use for autoresearch or "optimize X overnight/in a loop". Plan bounded, measurable experiment batches, then delegate execution through orchestration.
Compare agent guidance, workflows, and dependencies across repositories, then adopt the practices each recipient would benefit from.
Audit multiple interdependent repositories for alignment, then plan and apply surgical fixes for drift and duplication.
Create one decision-complete task handoff in its repository or on the Desktop for cross-repository work, return a command for a fresh interactive Codex session, and recommend Codex subagent delegation for complex tasks.
Redesign the assigned repository or repositories for agent understanding, control, and cumulative learning. Use when asked to make a system agent-intuitive, agent-ergonomic, or agent-accretive, then revise its design documents and plans.
Maintain README.md, AGENTS.md/CLAUDE.md, skills, and repo context. Continuously maintain repo-local skills when task evidence warrants deleting obsolete skills, merging overlapping skills, or creating a reusable skill.
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable, AI-navigable, or agent-legible, choose which invariants to…