naman10parikh/agentgrid / skills/agentgrid-orchestrate
agentgrid-orchestrate
VerifiedHand out and track work across a running agentgrid. Send a prompt to one pane, inject a long mission file, broadcast to every pane, read a pane's output and watch live status. Use after a grid is running, when the user says "send this to pane 3", "broadcast", "inject the mission", "what is pane 2 doing" or "monitor the grid".
agentgrid-orchestrate: hand out and track work
When to use
- A grid is already running (created with
agentgrid ROWSxCOLS, or the MCP toolgrid_create). - Each pane needs its own mission, or every pane needs the same instruction.
- You need to know which agents are working, waiting or done, without switching windows.
Steps
-
List the panes.
agentgrid status --jsonreturns{"panes":[{"ref","status","label","command"}]}. Use theref(for exampleagentgrid:1.3) or a pane id such as%42as the target. -
Send one prompt to one pane:
agentgrid send %42 "You are Builder 1. Implement the login form in src/auth/." -
Inject a long mission (a Markdown file) into one pane:
agentgrid inject %42 --file missions/builder-1.md -
Broadcast the same line to every pane in the current window:
agentgrid broadcast "Pull main, run the tests, then report what failed." -
Read and watch.
agentgrid read %42 -n 40prints the pane's last 40 lines.agentgrid monitorshows every pane's status and recent output, refreshed live.agentgrid statusgives a one-shot summary. -
Follow-ups go to the owner. When a pane's work produces a follow-up bug, send it to that same pane with
agentgrid send. Do not broadcast it: the owner already holds the context.
From another coding agent, the same steps are MCP tools: grid_status, pane_send, pane_inject,
grid_broadcast, pane_read and grid_dashboard (run agentgrid-mcp, or install the plugin).
Output
Every pane has its mission, and you can report each pane's state. Report back to the user: which mission went to which pane, and the latest status of each (WORKING, WAITING, DONE or IDLE).
Do not
- Never
tmux select-pane: it steals the user's focus. - Never inject missions before the panes are named: names are how you and the user tell the work apart.
- Never send a secret (a key, a token) in a prompt: agents read keys from environment variables.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 469k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 321k | Updated | Claude CodeCodex | Verified | Install |