Orchestration

Ask the Cursor Composer agent read-only questions for a second opinion

Hand a single file edit to Cursor Composer in a worktree and review its diff

Run long tasks in a loop that re-enters with fresh context on each wake-up

Execute Plans.md tasks one at a time or as a full parallel team run

Keep going on reversible work instead of stopping to ask permission

Keep a TSV decision log so reviewers can trust long unattended runs

Fan out parallel workers and merge their results into one report

Run measured Git experiments toward a numeric target, keeping gains and reverting losses

Create and dispatch atomic work packets that need approval before any side effect

Run long, multi-step work with durable state, evidence and recovery

Coordinate teams of coding agent sessions with shared tasks and file locks

Split large migrations and refactors into parallel agent runs in separate worktrees

Run several LLMs in a council that ranks answers and synthesizes one result

Wire commands, agents and skills into research, plan and build phases

Track parallel coding sessions with status lines that name project, branch and task

Create webhook subscriptions so GitHub, Stripe or Sentry events trigger agent runs

Run a decision through five AI advisors that peer review each other for a verdict

Run code tasks across Codex, Claude, Gemini and OpenCode backends in parallel

Run long multi-session agent tasks with checkpoints, recovery and dependencies

Pick a minimal set of agents to analyze bugs, plan fixes and implement changes

Run a multi-agent team organized like Ming dynasty cabinet government

Supervise a long-running automated security review campaign across its stages

Design a custom deterministic harness that runs Mantis skills reliably

Delegate code analysis, refactors and automated edits to the Codex CLI

Run Gemini CLI for code review, plan review and very large context analysis

Delegate a coding task to the Antigravity agy CLI and review its diff before landing it

Delegate a coding task to Aider and review its diff before landing it

Hand a coding task to a separate Claude Code process, then review its diff

Delegate a coding task to the Cline CLI and review its diff before landing it

Run Codex delegations in the background without polling for progress

Delegate a coding task to the OpenAI Codex CLI and review its diff before landing it

Delegate a coding task to Command Code CLI and review its diff before landing it

Delegate a coding task to GitHub Copilot CLI and review its diff before landing it

Delegate a coding task to Cursor Agent CLI and review its diff before landing it

Delegate a coding task to the Grok Build CLI and review its diff before landing it

Delegate a coding task to the Kimi Code CLI and review its diff before landing it

Delegate a coding task to the Muse Code CLI and review its diff before landing it

Delegate a coding task to Oh My Pi and review its diff before landing it

Delegate a coding task to the OpenCode CLI and review its diff before landing it

Delegate a coding task to the Pi coding agent CLI and review its diff before landing it

Delegate a coding task to the Qoder CLI and review its diff before landing it

Delegate a coding task to Mistral Vibe CLI and review its diff before landing it

Delegate a coding task to the Warp Agent CLI and review its diff before landing it

Delegate a coding task to the Z.AI ZCode CLI and review its diff before landing it

Follow the circles protocol for connections, presence, messages and asks

Run coding agent CLIs like Codex and OpenCode as background processes

Check agent bounties, earnings, marketplace skills and network reputation

Parse YAML workflow files and run their multi-role steps in DAG order

Design orchestrator and subagent setups with exclusive write areas and audits